- Email: [email protected]
A simple color concentration measurement technique for powders
Powder Technology 286 (2015) 392–400
Contents lists available at ScienceDirect
Powder Technology journal homepage: www.elsevier.com/locate/powtec
A simple color concentration measurement technique for powders Heather N. Emady a,b, Maya Wittman b, Sara Koynov b, William G. Borghard b, Fernando J. Muzzio b, Benjamin J. Glasser b, Alberto M. Cuitino c,⁎ a b c
School for Engineering of Matter, Transport and Energy, Arizona State University, Tempe, AZ 85287, USA Department of Chemical and Biochemical Engineering, Rutgers University, Piscataway, NJ 08854, USA Department of Mechanical and Aerospace Engineering, Rutgers University, Piscataway, NJ 08854, USA
a r t i c l e
i n f o
Article history: Received 18 April 2015 Received in revised form 22 July 2015 Accepted 24 July 2015 Available online 20 August 2015 Keywords: Spectrophotometer Tracer Mixing Color Powder Calibration
a b s t r a c t Looking for a way to measure residence time distributions of an 80 micron fluidized cracking catalyst (FCC) powder, a simple measurement technique was discovered that quantifies tracer color concentration. Using a color spectrophotometer that measures percent reflectance as a function of wavelength, a calibration curve can be constructed for standard mixtures of dyed and un-dyed powder. This calibration curve can then be used to determine the color concentration of an unknown sample by measuring its reflectance. The effects of operating parameters such as dye strength, aperture size, surface roughness, sample volume and depth, and continuous flow were all evaluated. This spectrophotometric technique was found to be a quick and simple way to measure colored mixture concentrations. In addition to being ideal for residence time distribution applications, it has the potential to easily quantify mixing in any unit operation, batch or continuous. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Powder processing is prevalent in many industries, and is embodied in various unit operations [1–3]. Often, it is desirable to characterize a process via the addition of a tracer. Concentration of the tracer is usually the desired measurement in order to determine process characteristics such as residence time, residence time distribution, axial dispersion, and mixing [4–6]. A tracer can be a portion of the bulk material that is dyed, or it can be an entirely different material. Unless the system is inherently multicomponent, it is ideal to dye a portion of the original material as a tracer. Using the same material reduces potential inconsistencies that may arise due to differing material and flow properties of the bulk powder and tracer. Nonetheless, it is important to ensure that dyeing does not affect particle properties. The motivation for this work stems from the desire to measure residence time distributions of a powder in a rotary calciner, following a similar protocol that Gao et al. used for millimetric particles [7]. These researchers measured color concentration by visually counting the number of dyed particles in images, which is not practical for the greater numbers of particles that arise with a fine powder. As the particle size decreases, individual particles become more difficult to distinguish from one another, making image analysis exceedingly challenging. In order for colored tracers of various particle sizes to be successfully implemented, there needs to be a robust way to characterize the ⁎ Corresponding author. E-mail address: [email protected] (A.M. Cuitino).
http://dx.doi.org/10.1016/j.powtec.2015.07.050 0032-5910/© 2015 Elsevier B.V. All rights reserved.
concentration of tracer. The existing methods of measuring colored tracer concentration include image analysis and color spectrophotometry, which are detailed in the following sections.
1.1. Image analysis Image analysis for color concentration determination typically involves taking images of the multicolored particulate mixtures with a camera, and performing various post-processing steps on the images to extract the solid fraction. Although the particles are colored, most image acquisition and analysis is in black and white. Through image analysis, some researchers correlated gray scale values to tracer concentration [8,9]. Grasa and Abanades found a logarithmic relationship between the gray scale values in an image to the solid concentration, using white 0.85 mm PVC particles mixed with 0.5 mm coal in a fluidized bed [8]. Realpe and Velazquez correlated image gray scale values with powder concentration of different binary combinations of lactose, chocolate, and cellulose particles 44–90 μm in size in static powder beds [9]. With this method, they produced calibration curve fits by second, third, and fifth order polynomials, depending on the powder combination, all with R2 values above 0.995. Other authors used image analysis to asses mixing by taking images from a camera placed above a continuous conveyor belt setup containing the outflow of a mixer [10,11]. Muerza et al. studied a binary system of white aspirin and yellow semolina, and analyzed the images via an auto-correlation method [10]. Berthiaux et al. implemented principal component analysis to
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
determine a homogeneity ratio from images of white and black semolina particles 100–250 μm in size [11]. Some authors have explored the acquisition and subsequent analysis of images with color, with the analysis performed in RGB (red, green, and blue) color space. Aissa et al. pointed out that gray scale image analysis cannot be implemented for a mixture of more than two colored components due to overlap, resulting in the necessity for RGB image analysis [12]. They studied the mixing of red, yellow, blue, and white linear medium density polyethylene powders 83–550 μm in diameter in a rotating cylinder, analyzing the RGB images via multivariate image analysis of the pixel color intensities. Also implementing RGB analysis, the SolidSizer (JM Canty Inc., Buffalo, NY) measures size, shape, and color characteristics from images it takes of a stream of falling particles [13]. Using this instrument, Langroudi et al. studied white particles with a median size of 453 μm, and dyed some of these particles red to create the tracer. Here, the authors were able to distinguish the red particles as those having a larger R value relative to the sum of all of the color values (i.e., R + G + B, the total brightness). They measured tracer concentration as a function of time, allowing for the visualization of the residence time distribution of an axial flow Couette device [13]. Although the aforementioned researchers were successful in using image analysis, they employed many intensive post-processing and analysis steps. Attempting to replicate these methods is often not straightforward and cannot be accomplished simply by reading the relevant manuscript, resulting in time consuming efforts that may not be fruitful. Besides the image processing procedures, environmental factors such as lighting must be just right in order to capture quality images in the first place. Another point to consider is that, in each case, specific protocols were developed for a given particulate system that may not work for formulations outside the studied particle size range. Thus, there is not a single image analysis procedure that can be successfully applied in any situation. 1.2. Color spectrophotometry Color spectrophotometers, or colorimeters, are instruments that shine light on a sample and measure reflectance as a function of wavelength, and usually report color information in CIE color space. Most users tend to work with the CIELAB (Cartesian coordinates) or CIELCH (cylindrical coordinates) color information, where L* characterizes white and black, a* characterizes red and green, b* characterizes yellow and blue, C* characterizes hue intensity, and h characterizes hue angle [14–17]. CIELCH can be obtained from CIELAB via a mathematical transformation [17]. The spectrophotometry/CIE method seems to be employed often in the food and agricultural industries, where color can indicate freshness, ripeness, and in general whether or not a substance is visually pleasing [14,16]. As a direct measurement of substance color, McCaig measured L*a*b* values of 50 different particulate food samples using two different categories of instruments, colorimeters/spectrophotometers and visible near-infrared (VNIR) reflectance instruments [14]. They used a tristimulus colorimeter (CR-310, Minolta Canada Inc., Mississauga, ON) with a granular material attachment that extracts color information from a 50 mm diameter circle of the sample, as well as two hand-held spectrophotometers (CM-525i Minolta Canada Inc., Mississauga, ON), which use a 25 mm diameter circle of the sample. McCaig compared these direct L*a*b* color measurements to those calculated from spectra taken from the VNIR instruments. Three different VNIR NIRSystems 6500 instruments (Foss North America, Eden Prairie, MN) were used, which measure reflectance of a 36 mm diameter circle sample in the 400–2499 nm wavelength range. Reflectance measurements in the 400 to 780 nm wavelength range were used in the calculation of L*a*b* values from the VNIR data. In comparing these L*a*b* values to those directly obtained from the spectrophotometers, McCaig concluded that both types of instruments produce similar L*a*b* values. While colorimeters directly provide L*a*b* values, VNIR requires time
393
consuming manipulations to extract these values, but VNIR has the added advantage of providing chemical information from the NIR spectra. A few researchers used spectrophotometers to characterize powder mixtures [15–17]. Slettengren et al. used a portable tri-stimulus colorimeter (Chroma-Meter CR-300, Minolta AG, Dietikon, Switzerland) which analyzes an 8 mm diameter circle of the sample [15]. They used the a* values to calculate the coefficient of variation in order to assess mixing quality of powder–powder and powder–liquid systems. These systems comprised various combinations of palm stearin fat, sunflower oil, and two different types of flour, with particle sizes in the 1–200 μm range. Shenoy et al. applied the DigiEye (VeriVide Ltd., UK) digital color imaging system to assess mixing in food powders, examining binary mixtures of different combinations of salt (454 μm), black pepper (369 μm), paprika (252 μm), and onion (65 μm) [16]. This instrument has the capability to measure color in L*a*b*, as well as reflectance as a function of wavelength in the 400–700 nm range. From the L*a*b* values, the researchers calculated ΔE, which is the color value relative to white space, and characterized mixing quality by the variance in ΔE. However, they found that ΔE could not distinguish color concentrations above a certain color threshold, which was dependent upon the specific binary mixture (e.g., the ΔE for onion–salt increased until 30% onion, above which there was no relationship). Barling et al. used color tracers to assess mixing of pharmaceutical dry powder inhaler blends using CIELCH color space [17]. They studied blends of 1% sub-micronized iron oxide (red) and various grades of 99% lactose (white) in seven different mixers. This low amount of tracer did not appear to impact particle size distributions of the lactose. In each experiment, the authors sampled 5 g of material at different time points, and measured color with a colorimeter (ColorFlex EZ 45/0, HunterLab Inc., U.S.A.). Four measurements were taken of the sample at different rotations of a 6 cm diameter by 3.5 cm high sample cup and averaged. The authors found hue intensity (C*) to identify dispersion and hue angle (h) to identify de-agglomeration in powder mixing, and developed mixing curves based on these coordinates. From this information, a given formulation's mixing curve can be constructed, which can be used to assess mixing degree. Although the above researchers used spectrophotometers to measure color using the CIE system, we could not find any literature reporting the use of the raw spectral data, reflectance as a function of wavelength. Our interest in this spectral data stems from the desire to measure color concentration. With reflectance as a function of wavelength for standard tracer concentrations, a calibration curve can be constructed at a given wavelength. With a calibration curve of tracer concentration versus reflectance, the reflectance of unknown samples can be measured with the spectrophotometer, and their color concentrations determined. Thus, the technical gap that must be addressed is the analysis of spectral data from a color spectrophotometer for powder color concentration determination. In this paper, we present a straightforward way to measure color concentration of a powder mixture via spectral data from a spectrophotometer, which has not been demonstrated previously. Testing all of the different operational modes of a color spectrophotometer, we present our findings on the use of the fundamental spectral measurements, make recommendations on the configurations that provide optimal calibration curve fits, and apply the technique to an axial mixing experiment. 2. Materials and methods 2.1. Sample preparation The material used in this work is a fluidized cracking catalyst (FCC) powder supplied by Grace Davison (Columbia, MD). The tracer powder is created by dying the FCC powder with Sharpie ink. First, the top of the
394
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
Sharpie is broken off and the ink packet is removed. Acetone is used to rinse the ink from the top as well as the ink packet into a beaker. After the ink from the beaker is poured onto the powder, the acetone is used again to wash out all of the excess ink from the beaker. Approximately 30 mL of acetone is used per Sharpie. Following the addition of ink with acetone into the powder bed, the components are fully mixed together in the beaker using a glass stirring rod until the powder is consistently dyed. After one Sharpie's worth of ink is mixed with the powder completely, another is added. This process is repeated until the desired dye strength is achieved. Dye strength is measured by the number of Sharpies used to dye 100 mL of powder. Dye strengths of 3–6 dyeing procedures were prepared and tested. Samples of the dyed powder were used to create mixtures with undyed powder. For example, 0.05 g of colored powder mixed with 49.95 g of un-dyed powder forms a sample that is 0.1% colored by mass. Continuing with this process, mixtures of 0.1–1%, 1–10%, and 10–100% by mass dyed powder were generated with ten equal increments for each percentage range. This wide range of samples aids in determining the extent of variations in powder concentration that the measurement device can detect.
2.2. Effect of dyeing process on material properties The effect of the dyeing process on the material properties was determined by comparing the particle size distributions, conditioned bulk densities and compressibility indices, and SEM images of FCC dyed with 8 sharpies per 150 g to the un-dyed FCC. The particle size distributions were measured using a Beckman-Coulter LS 13 320 series laser diffraction particle size analyzer (Pasadena, CA, USA), and the d10, d50, and d90 of both the dyed and un-dyed FCC were 51, 89, and 139 μm, respectively. The particle size distributions are shown in Fig. 1a. The conditioned bulk densities and compressibility indices were measured using an FT4 powder rheometer (Freeman Technology, Tewkesbury, Gloucestershire, UK). The conditioned bulk density for the dyed FCC was 0.96 g/cm3, compared to that of the un-dyed FCC, 0.77 g/cm3. The compressibility index was computed as a ratio of the density at an applied normal stress of 15 kPa to the conditioned bulk density, and the compressibility index for the dyed FCC was 1.03, compared to that of the un-dyed FCC, which was 1.02. The compressibility profiles are shown in Fig. 1b. Some difference between the curves can be observed; however, the magnitude of the difference is smaller than the difference between free-flowing and poorly flowing materials. The compressibility percent of poorly flowing materials such as magnesium stearate can reach 35%. SEM photographs were taken of the dyed and un-dyed material to determine any effect on particle shape, shown in Fig. 2. It is observed that dyeing does not change the particle shape; however, the dyed material surface appears to be slightly smoother than that of the un-dyed material.
From this extensive material property comparison, we conclude that the dyed FCC is not significantly different from the un-dyed FCC. Thus, these materials are only distinguished by color. 2.3. Measurement method We use a color spectrophotometer (X-Rite VeriColor Spectro 450, Grand Rapids, MI) to determine the variation in color concentration between samples. Each sample is poured out evenly and smoothly into the sample spoon (as shown in Fig. 3a) and inserted in the machine to be scanned (see Fig. 3d). To see how different amounts and depths of powder are measured in the color probe, we use three different spoon sizes, as shown in Fig. 3b. From bottom to top, first is the small spoon, which holds approximately 3.2 mL of powder, with a depth of 9 mm. Next, the partially filled big spoon holds 5.4 mL of powder and has a depth of 4 mm. Finally, the filled big spoon has a volume of 21.5 mL and a depth of 14 mm. Additionally, there are two options for the aperture size, 12 mm and 6 mm. The small spoon can only be used with the 6 mm aperture. Although all spoon and aperture sizes were tested, our standard case consists of the partially filled (5.4 mL) spoon and the 12 mm aperture. Variations on this standard are noted when applicable. In addition to spoon and aperture sizes, other system variables include the measurement time, sample surface smoothness, and continuous operation. The measurement time is 750 ms, and there is a 1 s time interval between measurements. Usually, before inserting a sample spoon into the color probe, the surface is scraped level. As a way to see how the results would change if this was not the case, the powder is poured into the spoon without the leveling step, creating a rough surface, and then the measurements are taken. Continuous operation is possible in 5 s to 1 h intervals. To test this mode, the spectrophotometer was placed over a conveyer belt (width of 15 cm, length of 88 cm) of powder running at its lowest speed, where the instrument automatically recorded measurements every few seconds (see Fig. 3e). The spectrophotometer was placed on a stand that also leveled the powder on the conveyor belt, both creating a smooth surface and the correct distance for measurement acquisition. Four color measurements are performed for each sample. The first scans the powder with the spoon fully inserted in the spectrophotometer. The next three are “trials” (see Fig. 3c), with the first having the spoon fully in (target), the second halfway in (middle), and the third is just the tip of the spoon (tip). These measurements show if there is any difference in the data based on which part of the powder sample is scanned. In order to find possible variations in scanning the same sample, measurements are repeated by pouring the sample out, mixing it, and pouring it back into the spoon. 2.4. Measurement outcomes The spectrophotometer collects data by shining a full spectrum LED light on the sample, and measures percent reflectance as a function of
Fig. 1. (a) Particle size distributions and (b) compressibility profiles for un-dyed and dyed FCC.
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
395
Fig. 2. SEM photographs of (a) un-dyed and (b) dyed FCC.
3. Results and discussion
wavelength in the 350–750 nm range. In order to draw conclusions about this data, we graph percent reflectance versus wavelength, as shown in Fig. 4. Each sample has a different percentage of dyed powder and is shown on the graph as a different curve of uniquely-shaped data points. As the concentration of dyed powder increases, the percent reflectance of the sample decreases. Thus, the lighter powders with the smallest concentrations of dyed powder have a higher percent reflectance at each corresponding wavelength.
Here, the effects of different measurement operational variables on the calibration curves are compared. In all cases, the calibration curves were fit with fourth order polynomials with R2 values of at least 0.997. However, variations become discernable in the 0–10% dyed powder range. Therefore, comparisons will be made only using this low dyed powder concentration range.
2.5. Calibration curve development
3.1. Dye strength
In any tracer experiment, a calibration curve is needed to relate the instrument output to the color concentration of the sample. Taking the data from a single wavelength (570 nm), it is possible to graph the percent dyed powder versus the percent reflectance (see Fig. 5). This wavelength was chosen because it showed the lowest standard deviation across each set of samples. The most consistent data will provide the best curve fit, which can ideally give predictions of a sample with an unknown concentration based on its percent reflectance.
To determine the optimum dye strength, we analyzed the differences in spectra retrieved from powder mixtures containing various amounts of dye, from 3–6 dyeing procedures. From the calibration curves in the 1–10% dyed powder range in Fig. 6, it is apparent that the data for 5 dyeing procedures (see Fig. 6c) follows the most distinct pattern and has the best curve fit. Therefore, the optimum dye strength for this material was achieved by using 5 Sharpies per 100 mL of powder.
Fig. 3. (a) Sample filled with powder mixture of equal parts dyed and un-dyed powder; (b) Three spoon sizes of volumes 3.2 mL, 5.4 mL, and 21.5 mL, from bottom to top; (c) 21.5 mL sample spoon showing the three different trials; (d) Spoon getting scanned in the spectrophotometer with the 12 mm aperture; and (e) Spectrophotometer placed above a conveyor belt of powder for operation in continuous mode.
396
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
Fig. 4. Spectra for 5 dyeing procedures. Graph shows percent reflectance as a function of wavelength for dyed mixture concentrations of 0.1–1%, 1–10% and 10–100%, each in increments of 10. Percent reflectance decreases as the concentration of dyed powder increases. The legend indicates the percent dyed amount of the sample.
3.2. Aperture size and surface roughness There are two options for aperture size, 6 mm and 12 mm. Although the small spoon can only handle the 6 mm aperture size, there is an option to use either aperture size with the large spoon. To test the effect of aperture size, each aperture size was tested on samples in the large (21.5 mL) spoon. Additionally, a rough surface was prepared and tested with the 12 mm aperture to ascertain the sensitivity of the measurement to surface roughness. Table 1 shows the calibration curves and corresponding R2 values for the 12 mm aperture and smooth surface (top), 6 mm aperture and smooth surface (middle), and 12 mm aperture and rough surface (bottom). From the top and middle lines in the table, taking both the curves and the R2 values in conjunction, the 12 mm aperture gives a better curve fit than the 6 mm aperture. This is expected, as the large aperture captures a larger sample area, which is more representative of the entire powder sample. Looking at the top and bottom lines in Table 1, it is apparent that a rough surface provides a poor curve fit, with an R2 of 0.869. Here, when the concentration of dyed powder is less than 1%, the percent reflectance varies widely from around 64% to 77%, for a span of about 13%, while this span is less than 3% for the smooth surface. This variation with the rough surface is expected, as an uneven surface makes the measurement less reproducible. Based on these observations, it is recommended to always use the 12 mm aperture with the large spoon, and to always ensure the sample surface is smooth.
in depth) and a large spoon (21.5 mL, 14 mm). A modification was made to the large spoon by placing a piece of cardboard inside, thus forming the partially filled big spoon (5.4 mL, 4 mm). Usually, the decision on spoon size is based upon the quantity of sample; the large spoon is used if large amounts are available, and the small spoon is used when the material quantity is limited. Now, the partially filled big spoon is an alternative option regardless of the material sample size. The partial spoon also allows for smaller samples to be scanned with the larger (12 mm) aperture. The 12 mm aperture was used for the partial and large spoons, while the 6 mm aperture was used for the small spoon. Table 2 shows the calibration curves for the small spoon (top), partial large spoon (middle), and large spoon (bottom). The large spoon, with the largest volume and depth, appears to provide the best curve fit, with an R2 of 0.997. The worst curve fit is for the small spoon, with an R2 of 0.983, while the partially filled large spoon falls in between with an R2 of 0.991. Based on these curve fits, it is recommended to use the large spoon when the quantity of sample is sufficient. When only small sample quantities are available, the partially filled large spoon may be a better option than the small spoon, since the former can be scanned with the 12 mm aperture. 3.4. Continuous flow Only white, un-dyed powder was tested in the case of continuous flow, to illustrate the general feasibility of continuous measurements with this instrument. Two different continuous runs were performed, one using the 12 mm aperture, and one with the 6 mm aperture. The resulting spectra are given in Fig. 7, where each curve indicates a measurement taken at a given elapsed time. From these results, it appears that the spectra for the 12 mm aperture vary over time, while the spectra for the 6 mm aperture are more consistent aside from one outlier time point. The primary cause of variation in continuous measurements is likely to be the surface of the powder bed. Although our setup included a scraping mechanism by the instrument stand, it is difficult to ensure a completely smooth surface. Unlike in static measurements, where the vertical distance is fixed in place by the sample spoon holder, in continuous mode, it is difficult to ensure that the smoothness and correct vertical distance remain after being scraped, as the powder still must travel a short horizontal distance before the measurement is taken. Thus, the 6 mm aperture may be a better option for continuous mode since it captures a smaller area, and therefore will not be as sensitive to macroscopic smoothness issues on the powder surface. However, we recommend that both aperture sizes be tested for any continuous mode setup, and care should be taken to ensure the powder surface is as smooth as possible.
3.3. Sample volume and depth 4. Axial mixing case study Three different sample sizes were tested that varied in volume and depth. The device comes with a small spoon (3.2 mL in volume, 9 mm
The horizontal rotating drum is a commonly used unit operation across industries, including calcination in catalytic manufacturing. For batch operations, the drum or cylinder is placed in a horizontal position and rotated about its axis. For continuous operations, the drum is often arranged in an inclined configuration, typically by a few degrees. Calciners, along with other unit operations, are often characterized by their residence time distribution (RTD) [18]. The RTD can be related to the axial dispersion coefficient using the Taylor dispersion model and Peclet number [19,20]: Peð1−θÞ2 1 EðθÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi e− 4θ 2 πθ=Pe
Fig. 5. Calibration curve for 5 dyeing procedures. The graph shows the percent dyed powder as a function of percent reflectance at a wavelength of 570 nm. The curve was fit with a fourth order polynomial, with an R2 of 0.99975.
where E(θ) is the residence time distribution, θ is the dimensionless time, and Pe is the Peclet number (the ratio of advective to dispersive transport). This approach can be used to characterize continuous systems, and so would be most applicable to pilot-scale systems. To aid in the design and scale-up of calciners, measuring the axial dispersion
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
397
Fig. 6. Calibration curves in the 1–10% dyed powder range, for (a) 3 dyeing procedures, (b) 4 dyeing procedures, (c) 5 dyeing procedures, and (d) 6 dyeing procedures. Fourth order polynomial fits provide R2 values of 0.974, 0.964, 0.984, and 0.968, respectively.
Table 1 Comparison of the effects of aperture size and surface roughness on the calibration curve fits. Calibration curve fit with fourth order polynomial
R2
Aperture size (mm)
Surface (smooth or rough)
12
Smooth
0.997
6
Smooth
0.991
12
Rough
0.869
398
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
Table 2 Comparison of the effects of sample volume and depth on the calibration curve fits. Sample volume (mL)
Sample depth (mm)
Calibration curve fit with fourth order polynomial
R2
3.2
9
0.983
5.4
4
0.991
21.5
14
0.997
coefficient in a bench-scale closed (i.e., non-continuous) system would be advantageous [21]. Using the Peclet number, an axial dispersion coefficient measured in a non-continuous system can be compared to that measured in a continuous flow system. This approach is illustrated using a case study below. It has been found that the axial dispersion of particulate systems can be described using Fick's law [22]:
2
∂c ∂ c ¼ Dax 2 ∂t ∂x
where c is the concentration dyed material, t is the time, x is the axial position, and Dax is the axial dispersion coefficient. In this study, a cylinder 4 cm in diameter and 30 cm in length was loaded with un-dyed FCC in the first 10 cm, dyed FCC in the second 10 cm, and un-dyed FCC in the third 10 cm, to a fill level of 35%. The cylinder was rotated at 25 rpm for 30 min. At 5 min intervals, the cylinder was stopped and 2.5 mL samples were taken every 3 cm. The color spectrophotometer was used to measure the reflectance of each sample (see Fig. 8a), and a calibration curve was used to determine the concentration of dyed material in each sample (see Fig. 8b). Concentration profiles were created as a function of axial position for each time point. Fick's second law was solved
Fig. 7. Spectra for un-dyed powder flowing continuously on a conveyor belt, measured with the (a) 12 mm aperture and (b) 6 mm aperture.
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
399
Fig. 8. (a) Percent reflectance of samples of dyed and un-dyed FCC. (b) Concentration of dyed material in FCC samples calculated using a calibration curve.
numerically, and the axial dispersion coefficient that minimized the residual between the numerical solutions was identified. For three replicates of the six time points, the average value was 0.016 cm2/min with a standard deviation of 0.007 cm2/min. This value is in agreement with axial dispersion coefficients of powders previously found in the literature [21,23]. 5. Conclusions A simple color measurement system that measures reflectance as a function of wavelength was successfully applied to determine tracer powder color concentration. Calibration curves were constructed at a single wavelength to obtain color concentration as a function of reflectance. The measurements were the most sensitive to concentrations less than 10% dyed material, so this was the region used for calibration curve comparisons. All of the system operating conditions were explored to ascertain their effects on calibration curves. Dye strength, aperture size, surface roughness, sample volume and depth, and continuous flow were all tested. Preparing mixtures of varying tracer concentration, static measurements were performed to obtain spectral color information. Five dyeing procedures (5 Sharpies/100 mL powder) was the ideal dye strength for our 80 micron FCC material. The best sample chamber and aperture was the large spoon with the 12 mm aperture. When only a limited amount of material is available, the partial large spoon may be better than the small spoon since the former can be scanned with the larger aperture. A partial large spoon can easily be constructed by the insertion of a flat solid object, such as a piece of cardboard, into the large spoon. With every sample, care should be taken to ensure the sample surface is as smooth as possible. The feasibility of using the spectrophotometer for continuous flow was explored. Both aperture sizes were tested on un-dyed powder flowing continuously on a conveyor belt, with measurements taken every few seconds. Here, the smaller 6 mm aperture may be better because it is likely to be less sensitive to macroscopic surface roughness. Continuous operation will need to be configured for each application, but this setup has the potential to be used as an in-line color tracer measurement method. Finally, we applied our color spectrophotometric technique to an axial dispersion experiment of a mixture of dyed and un-dyed powder in a rotating cylinder. Taking samples across the cylinder over time and analyzing the resulting spectra validated that our method can easily be applied to assess mixing. Although a specific color spectrophotometer was used in these measurements, it is expected that the general recommendations will translate to other colorimeters that measure reflectance as a function of wavelength. For future work, it would be useful to test the particle size limits of the spectrophotometry technique. As one can imagine, as particle size increases, the natural packing fraction decreases, meaning individual particles may start to protrude and produce a more uneven surface. As was shown in the present study, surface roughness causes
non-reproducible calibration curves. It is expected that there is a particle size threshold below which spectrophotometry would be more effective, and above which image analysis may be the better option. Returning to the original motivation for this work, this spectrophotometric technique is ideal for measuring residence time distributions (RTDs). Samples can be collected at the outlet of the unit operation and measured off-line, or the spectrophotometer can be placed directly above the flowing material to take continuous measurements in-line. As it is a simple method that can be applied to both batch and continuous systems, this spectrophotometric measurement technique has the potential to transform the way color concentrations of powder mixtures are quantified. Acknowledgments The authors would like to acknowledge funding from the Rutgers Catalyst Manufacturing Science and Engineering Consortium. The authors would also like to thank Dr. James Scicolone for taking the SEM photographs. References [1] S.L. Lee, T.F. O'Connor, X. Yang, C.N. Cruz, S. Chatterjee, R.D. Madurawe, et al., Modernizing pharmaceutical manufacturing: from batch to continuous production, J. Pharm. Innov. (2015), http://dx.doi.org/10.1007/s12247-015-9215-8. [2] M. Poux, P. Fayolle, J. Bertrand, D. Bridoux, J. Bousquet, Powder mixing: some practical rules applied to agitated systems, Powder Technol. 68 (1991) 213–234, http://dx.doi.org/10.1016/0032-5910(91)80047-M. [3] E.M. Holt, The properties and forming of catalysts and absorbents by granulation, Powder Technol. 140 (2004) 194–202, http://dx.doi.org/10.1016/j.powtec.2004.01. 010. [4] Y. Gao, F.J. Muzzio, M.G. Ierapetritou, A review of the residence time distribution (RTD) applications in solid unit operations, Powder Technol. 228 (2012) 416–423, http://dx.doi.org/10.1016/j.powtec.2012.05.060. [5] S. Tallon, C.E. Davies, B. Barry, Slip velocity and axial dispersion measurements in a gas-solid pipeline using particle tracer analysis, Powder Technol. 99 (1998) 125–131, http://dx.doi.org/10.1016/S0032-5910(98)00095-3. [6] X. Deng, J. Scicolone, X. Han, R.N. Davé, Discrete element method simulation of a conical screen mill: a continuous dry coating device, Chem. Eng. Sci. 125 (2015) 58–74. [7] Y. Gao, B.J. Glasser, M.G. Ierapetritou, A. Cuitino, F.J. Muzzio, J.W. Beeckman, et al., Measurement of residence time distribution in a rotary calciner, AIChE J. 59 (2013) 4068–4076, http://dx.doi.org/10.1002/aic.14175. [8] G. Grasa, J.C. Abanades, A calibration procedure to obtain solid concentrations from digital images of bulk powders, Powder Technol. 114 (2001) 125–128, http://dx.doi. org/10.1016/S0032-5910(00)00262-X. [9] A. Realpe, C. Velázquez, Image processing and analysis for determination of concentrations of powder mixtures, Powder Technol. 134 (2003) 193–200, http://dx.doi. org/10.1016/S0032-5910(03)00138-4. [10] S. Muerza, H. Berthiaux, S. Massol-Chaudeur, G. Thomas, A dynamic study of static mixing using on-line image analysis, Powder Technol. 128 (2002) 195–204, http://dx.doi.org/10.1016/S0032-5910(02)00197-3. [11] H. Berthiaux, V. Mosorov, L. Tomczak, C. Gatumel, J.F. Demeyre, Principal component analysis for characterising homogeneity in powder mixing using image processing techniques, Chem. Eng. Process. Process Intensif. 45 (2006) 397–403, http://dx.doi. org/10.1016/j.cep.2005.10.005. [12] A. Aït Aissa, C. Duchesne, D. Rodrigue, Polymer powders mixing part II: multicomponent mixing dynamics using RGB color analysis, Chem. Eng. Sci. 65 (2010) 3729–3738, http://dx.doi.org/10.1016/j.ces.2010.03.007.
400
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
[13] M. Kheiripour Langroudi, P.R. Mort, G.I. Tardos, Study of powder flow patterns in a Couette cell with axial flow using tracers and solid fraction measurements, Granul. Matter 13 (2011) 541–552, http://dx.doi.org/10.1007/s10035-011-0282-3. [14] T.N. McCaig, Extending the use of visible/near-infrared reflectance spectrophotometers to measure colour of food and agricultural products, Food Res. Int. 35 (2002) 731–736, http://dx.doi.org/10.1016/S0963-9969(02)00068-6. [15] K. Slettengren, P. Heunemann, O. Knuchel, E.J. Windhab, Mixing quality of powder–liquid mixtures studied by near infrared spectroscopy and colorimetry, Powder Technol. 278 (2015) 130–137, http://dx.doi.org/10.1016/j.powtec. 2015.03.020. [16] P. Shenoy, F. Innings, T. Lilliebjelke, C. Jonsson, J. Fitzpatrick, L. Ahrné, Investigation of the application of digital colour imaging to assess the mixture quality of binary food powder mixes, J. Food Eng. 128 (2014) 140–145, http://dx.doi.org/10.1016/j. jfoodeng.2013.12.013. [17] D. Barling, D.A.V. Morton, K. Hapgood, Pharmaceutical dry powder blending and scale-up: maintaining equivalent mixing conditions using a coloured tracer powder, Powder Technol. 270 (2015) 461–469, http://dx.doi.org/10.1016/j.powtec.2014.04. 069.
[18] Y. Gao, A. Vanarase, F. Muzzio, M. Ierapetritou, Characterizing continuous powder mixing using residence time distribution, Chem. Eng. Sci. 66 (2011) 417–425, http://dx.doi.org/10.1016/j.ces.2010.10.045. [19] A.T. Harris, J.F. Davidson, R.B. Thorpe, The influence of the riser exit on the particle residence time distribution in a circulating fluidised bed riser, Chem. Eng. Sci. 58 (2003) 3669–3680, http://dx.doi.org/10.1016/S0009-2509(03)00215-X. [20] A.T. Harris, J.F. Davidson, R.B. Thorpe, Particle residence time distributions in circulating fluidised beds, Chem. Eng. Sci. 58 (2003) 2181–2202, http://dx.doi.org/10. 1016/S0009-2509(03)00082-4. [21] R.G. Sherritt, J. Chaouki, A.K. Mehrotra, L.A. Behie, Axial dispersion in the threedimensional mixing of particles in a rotating drum reactor, Chem. Eng. Sci. 58 (2003) 401–415, http://dx.doi.org/10.1016/S0009-2509(02)00551-1. [22] P.M.C. Lacey, Developments in the theory of particle mixing, J. Appl. Chem. 4 (2007) 257–268, http://dx.doi.org/10.1002/jctb.5010040504. [23] A.S. Bongo Njeng, S. Vitu, M. Clausse, J.-L. Dirion, M. Debacq, Effect of lifter shape and operating parameters on the flow of materials in a pilot rotary kiln: part I. Experimental RTD and axial dispersion study, Powder Technol. 269 (2015) 554–565, http://dx.doi.org/10.1016/j.powtec.2014.03.066.
Contents lists available at ScienceDirect
Powder Technology journal homepage: www.elsevier.com/locate/powtec
A simple color concentration measurement technique for powders Heather N. Emady a,b, Maya Wittman b, Sara Koynov b, William G. Borghard b, Fernando J. Muzzio b, Benjamin J. Glasser b, Alberto M. Cuitino c,⁎ a b c
School for Engineering of Matter, Transport and Energy, Arizona State University, Tempe, AZ 85287, USA Department of Chemical and Biochemical Engineering, Rutgers University, Piscataway, NJ 08854, USA Department of Mechanical and Aerospace Engineering, Rutgers University, Piscataway, NJ 08854, USA
a r t i c l e
i n f o
Article history: Received 18 April 2015 Received in revised form 22 July 2015 Accepted 24 July 2015 Available online 20 August 2015 Keywords: Spectrophotometer Tracer Mixing Color Powder Calibration
a b s t r a c t Looking for a way to measure residence time distributions of an 80 micron fluidized cracking catalyst (FCC) powder, a simple measurement technique was discovered that quantifies tracer color concentration. Using a color spectrophotometer that measures percent reflectance as a function of wavelength, a calibration curve can be constructed for standard mixtures of dyed and un-dyed powder. This calibration curve can then be used to determine the color concentration of an unknown sample by measuring its reflectance. The effects of operating parameters such as dye strength, aperture size, surface roughness, sample volume and depth, and continuous flow were all evaluated. This spectrophotometric technique was found to be a quick and simple way to measure colored mixture concentrations. In addition to being ideal for residence time distribution applications, it has the potential to easily quantify mixing in any unit operation, batch or continuous. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Powder processing is prevalent in many industries, and is embodied in various unit operations [1–3]. Often, it is desirable to characterize a process via the addition of a tracer. Concentration of the tracer is usually the desired measurement in order to determine process characteristics such as residence time, residence time distribution, axial dispersion, and mixing [4–6]. A tracer can be a portion of the bulk material that is dyed, or it can be an entirely different material. Unless the system is inherently multicomponent, it is ideal to dye a portion of the original material as a tracer. Using the same material reduces potential inconsistencies that may arise due to differing material and flow properties of the bulk powder and tracer. Nonetheless, it is important to ensure that dyeing does not affect particle properties. The motivation for this work stems from the desire to measure residence time distributions of a powder in a rotary calciner, following a similar protocol that Gao et al. used for millimetric particles [7]. These researchers measured color concentration by visually counting the number of dyed particles in images, which is not practical for the greater numbers of particles that arise with a fine powder. As the particle size decreases, individual particles become more difficult to distinguish from one another, making image analysis exceedingly challenging. In order for colored tracers of various particle sizes to be successfully implemented, there needs to be a robust way to characterize the ⁎ Corresponding author. E-mail address: [email protected] (A.M. Cuitino).
http://dx.doi.org/10.1016/j.powtec.2015.07.050 0032-5910/© 2015 Elsevier B.V. All rights reserved.
concentration of tracer. The existing methods of measuring colored tracer concentration include image analysis and color spectrophotometry, which are detailed in the following sections.
1.1. Image analysis Image analysis for color concentration determination typically involves taking images of the multicolored particulate mixtures with a camera, and performing various post-processing steps on the images to extract the solid fraction. Although the particles are colored, most image acquisition and analysis is in black and white. Through image analysis, some researchers correlated gray scale values to tracer concentration [8,9]. Grasa and Abanades found a logarithmic relationship between the gray scale values in an image to the solid concentration, using white 0.85 mm PVC particles mixed with 0.5 mm coal in a fluidized bed [8]. Realpe and Velazquez correlated image gray scale values with powder concentration of different binary combinations of lactose, chocolate, and cellulose particles 44–90 μm in size in static powder beds [9]. With this method, they produced calibration curve fits by second, third, and fifth order polynomials, depending on the powder combination, all with R2 values above 0.995. Other authors used image analysis to asses mixing by taking images from a camera placed above a continuous conveyor belt setup containing the outflow of a mixer [10,11]. Muerza et al. studied a binary system of white aspirin and yellow semolina, and analyzed the images via an auto-correlation method [10]. Berthiaux et al. implemented principal component analysis to
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
determine a homogeneity ratio from images of white and black semolina particles 100–250 μm in size [11]. Some authors have explored the acquisition and subsequent analysis of images with color, with the analysis performed in RGB (red, green, and blue) color space. Aissa et al. pointed out that gray scale image analysis cannot be implemented for a mixture of more than two colored components due to overlap, resulting in the necessity for RGB image analysis [12]. They studied the mixing of red, yellow, blue, and white linear medium density polyethylene powders 83–550 μm in diameter in a rotating cylinder, analyzing the RGB images via multivariate image analysis of the pixel color intensities. Also implementing RGB analysis, the SolidSizer (JM Canty Inc., Buffalo, NY) measures size, shape, and color characteristics from images it takes of a stream of falling particles [13]. Using this instrument, Langroudi et al. studied white particles with a median size of 453 μm, and dyed some of these particles red to create the tracer. Here, the authors were able to distinguish the red particles as those having a larger R value relative to the sum of all of the color values (i.e., R + G + B, the total brightness). They measured tracer concentration as a function of time, allowing for the visualization of the residence time distribution of an axial flow Couette device [13]. Although the aforementioned researchers were successful in using image analysis, they employed many intensive post-processing and analysis steps. Attempting to replicate these methods is often not straightforward and cannot be accomplished simply by reading the relevant manuscript, resulting in time consuming efforts that may not be fruitful. Besides the image processing procedures, environmental factors such as lighting must be just right in order to capture quality images in the first place. Another point to consider is that, in each case, specific protocols were developed for a given particulate system that may not work for formulations outside the studied particle size range. Thus, there is not a single image analysis procedure that can be successfully applied in any situation. 1.2. Color spectrophotometry Color spectrophotometers, or colorimeters, are instruments that shine light on a sample and measure reflectance as a function of wavelength, and usually report color information in CIE color space. Most users tend to work with the CIELAB (Cartesian coordinates) or CIELCH (cylindrical coordinates) color information, where L* characterizes white and black, a* characterizes red and green, b* characterizes yellow and blue, C* characterizes hue intensity, and h characterizes hue angle [14–17]. CIELCH can be obtained from CIELAB via a mathematical transformation [17]. The spectrophotometry/CIE method seems to be employed often in the food and agricultural industries, where color can indicate freshness, ripeness, and in general whether or not a substance is visually pleasing [14,16]. As a direct measurement of substance color, McCaig measured L*a*b* values of 50 different particulate food samples using two different categories of instruments, colorimeters/spectrophotometers and visible near-infrared (VNIR) reflectance instruments [14]. They used a tristimulus colorimeter (CR-310, Minolta Canada Inc., Mississauga, ON) with a granular material attachment that extracts color information from a 50 mm diameter circle of the sample, as well as two hand-held spectrophotometers (CM-525i Minolta Canada Inc., Mississauga, ON), which use a 25 mm diameter circle of the sample. McCaig compared these direct L*a*b* color measurements to those calculated from spectra taken from the VNIR instruments. Three different VNIR NIRSystems 6500 instruments (Foss North America, Eden Prairie, MN) were used, which measure reflectance of a 36 mm diameter circle sample in the 400–2499 nm wavelength range. Reflectance measurements in the 400 to 780 nm wavelength range were used in the calculation of L*a*b* values from the VNIR data. In comparing these L*a*b* values to those directly obtained from the spectrophotometers, McCaig concluded that both types of instruments produce similar L*a*b* values. While colorimeters directly provide L*a*b* values, VNIR requires time
393
consuming manipulations to extract these values, but VNIR has the added advantage of providing chemical information from the NIR spectra. A few researchers used spectrophotometers to characterize powder mixtures [15–17]. Slettengren et al. used a portable tri-stimulus colorimeter (Chroma-Meter CR-300, Minolta AG, Dietikon, Switzerland) which analyzes an 8 mm diameter circle of the sample [15]. They used the a* values to calculate the coefficient of variation in order to assess mixing quality of powder–powder and powder–liquid systems. These systems comprised various combinations of palm stearin fat, sunflower oil, and two different types of flour, with particle sizes in the 1–200 μm range. Shenoy et al. applied the DigiEye (VeriVide Ltd., UK) digital color imaging system to assess mixing in food powders, examining binary mixtures of different combinations of salt (454 μm), black pepper (369 μm), paprika (252 μm), and onion (65 μm) [16]. This instrument has the capability to measure color in L*a*b*, as well as reflectance as a function of wavelength in the 400–700 nm range. From the L*a*b* values, the researchers calculated ΔE, which is the color value relative to white space, and characterized mixing quality by the variance in ΔE. However, they found that ΔE could not distinguish color concentrations above a certain color threshold, which was dependent upon the specific binary mixture (e.g., the ΔE for onion–salt increased until 30% onion, above which there was no relationship). Barling et al. used color tracers to assess mixing of pharmaceutical dry powder inhaler blends using CIELCH color space [17]. They studied blends of 1% sub-micronized iron oxide (red) and various grades of 99% lactose (white) in seven different mixers. This low amount of tracer did not appear to impact particle size distributions of the lactose. In each experiment, the authors sampled 5 g of material at different time points, and measured color with a colorimeter (ColorFlex EZ 45/0, HunterLab Inc., U.S.A.). Four measurements were taken of the sample at different rotations of a 6 cm diameter by 3.5 cm high sample cup and averaged. The authors found hue intensity (C*) to identify dispersion and hue angle (h) to identify de-agglomeration in powder mixing, and developed mixing curves based on these coordinates. From this information, a given formulation's mixing curve can be constructed, which can be used to assess mixing degree. Although the above researchers used spectrophotometers to measure color using the CIE system, we could not find any literature reporting the use of the raw spectral data, reflectance as a function of wavelength. Our interest in this spectral data stems from the desire to measure color concentration. With reflectance as a function of wavelength for standard tracer concentrations, a calibration curve can be constructed at a given wavelength. With a calibration curve of tracer concentration versus reflectance, the reflectance of unknown samples can be measured with the spectrophotometer, and their color concentrations determined. Thus, the technical gap that must be addressed is the analysis of spectral data from a color spectrophotometer for powder color concentration determination. In this paper, we present a straightforward way to measure color concentration of a powder mixture via spectral data from a spectrophotometer, which has not been demonstrated previously. Testing all of the different operational modes of a color spectrophotometer, we present our findings on the use of the fundamental spectral measurements, make recommendations on the configurations that provide optimal calibration curve fits, and apply the technique to an axial mixing experiment. 2. Materials and methods 2.1. Sample preparation The material used in this work is a fluidized cracking catalyst (FCC) powder supplied by Grace Davison (Columbia, MD). The tracer powder is created by dying the FCC powder with Sharpie ink. First, the top of the
394
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
Sharpie is broken off and the ink packet is removed. Acetone is used to rinse the ink from the top as well as the ink packet into a beaker. After the ink from the beaker is poured onto the powder, the acetone is used again to wash out all of the excess ink from the beaker. Approximately 30 mL of acetone is used per Sharpie. Following the addition of ink with acetone into the powder bed, the components are fully mixed together in the beaker using a glass stirring rod until the powder is consistently dyed. After one Sharpie's worth of ink is mixed with the powder completely, another is added. This process is repeated until the desired dye strength is achieved. Dye strength is measured by the number of Sharpies used to dye 100 mL of powder. Dye strengths of 3–6 dyeing procedures were prepared and tested. Samples of the dyed powder were used to create mixtures with undyed powder. For example, 0.05 g of colored powder mixed with 49.95 g of un-dyed powder forms a sample that is 0.1% colored by mass. Continuing with this process, mixtures of 0.1–1%, 1–10%, and 10–100% by mass dyed powder were generated with ten equal increments for each percentage range. This wide range of samples aids in determining the extent of variations in powder concentration that the measurement device can detect.
2.2. Effect of dyeing process on material properties The effect of the dyeing process on the material properties was determined by comparing the particle size distributions, conditioned bulk densities and compressibility indices, and SEM images of FCC dyed with 8 sharpies per 150 g to the un-dyed FCC. The particle size distributions were measured using a Beckman-Coulter LS 13 320 series laser diffraction particle size analyzer (Pasadena, CA, USA), and the d10, d50, and d90 of both the dyed and un-dyed FCC were 51, 89, and 139 μm, respectively. The particle size distributions are shown in Fig. 1a. The conditioned bulk densities and compressibility indices were measured using an FT4 powder rheometer (Freeman Technology, Tewkesbury, Gloucestershire, UK). The conditioned bulk density for the dyed FCC was 0.96 g/cm3, compared to that of the un-dyed FCC, 0.77 g/cm3. The compressibility index was computed as a ratio of the density at an applied normal stress of 15 kPa to the conditioned bulk density, and the compressibility index for the dyed FCC was 1.03, compared to that of the un-dyed FCC, which was 1.02. The compressibility profiles are shown in Fig. 1b. Some difference between the curves can be observed; however, the magnitude of the difference is smaller than the difference between free-flowing and poorly flowing materials. The compressibility percent of poorly flowing materials such as magnesium stearate can reach 35%. SEM photographs were taken of the dyed and un-dyed material to determine any effect on particle shape, shown in Fig. 2. It is observed that dyeing does not change the particle shape; however, the dyed material surface appears to be slightly smoother than that of the un-dyed material.
From this extensive material property comparison, we conclude that the dyed FCC is not significantly different from the un-dyed FCC. Thus, these materials are only distinguished by color. 2.3. Measurement method We use a color spectrophotometer (X-Rite VeriColor Spectro 450, Grand Rapids, MI) to determine the variation in color concentration between samples. Each sample is poured out evenly and smoothly into the sample spoon (as shown in Fig. 3a) and inserted in the machine to be scanned (see Fig. 3d). To see how different amounts and depths of powder are measured in the color probe, we use three different spoon sizes, as shown in Fig. 3b. From bottom to top, first is the small spoon, which holds approximately 3.2 mL of powder, with a depth of 9 mm. Next, the partially filled big spoon holds 5.4 mL of powder and has a depth of 4 mm. Finally, the filled big spoon has a volume of 21.5 mL and a depth of 14 mm. Additionally, there are two options for the aperture size, 12 mm and 6 mm. The small spoon can only be used with the 6 mm aperture. Although all spoon and aperture sizes were tested, our standard case consists of the partially filled (5.4 mL) spoon and the 12 mm aperture. Variations on this standard are noted when applicable. In addition to spoon and aperture sizes, other system variables include the measurement time, sample surface smoothness, and continuous operation. The measurement time is 750 ms, and there is a 1 s time interval between measurements. Usually, before inserting a sample spoon into the color probe, the surface is scraped level. As a way to see how the results would change if this was not the case, the powder is poured into the spoon without the leveling step, creating a rough surface, and then the measurements are taken. Continuous operation is possible in 5 s to 1 h intervals. To test this mode, the spectrophotometer was placed over a conveyer belt (width of 15 cm, length of 88 cm) of powder running at its lowest speed, where the instrument automatically recorded measurements every few seconds (see Fig. 3e). The spectrophotometer was placed on a stand that also leveled the powder on the conveyor belt, both creating a smooth surface and the correct distance for measurement acquisition. Four color measurements are performed for each sample. The first scans the powder with the spoon fully inserted in the spectrophotometer. The next three are “trials” (see Fig. 3c), with the first having the spoon fully in (target), the second halfway in (middle), and the third is just the tip of the spoon (tip). These measurements show if there is any difference in the data based on which part of the powder sample is scanned. In order to find possible variations in scanning the same sample, measurements are repeated by pouring the sample out, mixing it, and pouring it back into the spoon. 2.4. Measurement outcomes The spectrophotometer collects data by shining a full spectrum LED light on the sample, and measures percent reflectance as a function of
Fig. 1. (a) Particle size distributions and (b) compressibility profiles for un-dyed and dyed FCC.
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
395
Fig. 2. SEM photographs of (a) un-dyed and (b) dyed FCC.
3. Results and discussion
wavelength in the 350–750 nm range. In order to draw conclusions about this data, we graph percent reflectance versus wavelength, as shown in Fig. 4. Each sample has a different percentage of dyed powder and is shown on the graph as a different curve of uniquely-shaped data points. As the concentration of dyed powder increases, the percent reflectance of the sample decreases. Thus, the lighter powders with the smallest concentrations of dyed powder have a higher percent reflectance at each corresponding wavelength.
Here, the effects of different measurement operational variables on the calibration curves are compared. In all cases, the calibration curves were fit with fourth order polynomials with R2 values of at least 0.997. However, variations become discernable in the 0–10% dyed powder range. Therefore, comparisons will be made only using this low dyed powder concentration range.
2.5. Calibration curve development
3.1. Dye strength
In any tracer experiment, a calibration curve is needed to relate the instrument output to the color concentration of the sample. Taking the data from a single wavelength (570 nm), it is possible to graph the percent dyed powder versus the percent reflectance (see Fig. 5). This wavelength was chosen because it showed the lowest standard deviation across each set of samples. The most consistent data will provide the best curve fit, which can ideally give predictions of a sample with an unknown concentration based on its percent reflectance.
To determine the optimum dye strength, we analyzed the differences in spectra retrieved from powder mixtures containing various amounts of dye, from 3–6 dyeing procedures. From the calibration curves in the 1–10% dyed powder range in Fig. 6, it is apparent that the data for 5 dyeing procedures (see Fig. 6c) follows the most distinct pattern and has the best curve fit. Therefore, the optimum dye strength for this material was achieved by using 5 Sharpies per 100 mL of powder.
Fig. 3. (a) Sample filled with powder mixture of equal parts dyed and un-dyed powder; (b) Three spoon sizes of volumes 3.2 mL, 5.4 mL, and 21.5 mL, from bottom to top; (c) 21.5 mL sample spoon showing the three different trials; (d) Spoon getting scanned in the spectrophotometer with the 12 mm aperture; and (e) Spectrophotometer placed above a conveyor belt of powder for operation in continuous mode.
396
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
Fig. 4. Spectra for 5 dyeing procedures. Graph shows percent reflectance as a function of wavelength for dyed mixture concentrations of 0.1–1%, 1–10% and 10–100%, each in increments of 10. Percent reflectance decreases as the concentration of dyed powder increases. The legend indicates the percent dyed amount of the sample.
3.2. Aperture size and surface roughness There are two options for aperture size, 6 mm and 12 mm. Although the small spoon can only handle the 6 mm aperture size, there is an option to use either aperture size with the large spoon. To test the effect of aperture size, each aperture size was tested on samples in the large (21.5 mL) spoon. Additionally, a rough surface was prepared and tested with the 12 mm aperture to ascertain the sensitivity of the measurement to surface roughness. Table 1 shows the calibration curves and corresponding R2 values for the 12 mm aperture and smooth surface (top), 6 mm aperture and smooth surface (middle), and 12 mm aperture and rough surface (bottom). From the top and middle lines in the table, taking both the curves and the R2 values in conjunction, the 12 mm aperture gives a better curve fit than the 6 mm aperture. This is expected, as the large aperture captures a larger sample area, which is more representative of the entire powder sample. Looking at the top and bottom lines in Table 1, it is apparent that a rough surface provides a poor curve fit, with an R2 of 0.869. Here, when the concentration of dyed powder is less than 1%, the percent reflectance varies widely from around 64% to 77%, for a span of about 13%, while this span is less than 3% for the smooth surface. This variation with the rough surface is expected, as an uneven surface makes the measurement less reproducible. Based on these observations, it is recommended to always use the 12 mm aperture with the large spoon, and to always ensure the sample surface is smooth.
in depth) and a large spoon (21.5 mL, 14 mm). A modification was made to the large spoon by placing a piece of cardboard inside, thus forming the partially filled big spoon (5.4 mL, 4 mm). Usually, the decision on spoon size is based upon the quantity of sample; the large spoon is used if large amounts are available, and the small spoon is used when the material quantity is limited. Now, the partially filled big spoon is an alternative option regardless of the material sample size. The partial spoon also allows for smaller samples to be scanned with the larger (12 mm) aperture. The 12 mm aperture was used for the partial and large spoons, while the 6 mm aperture was used for the small spoon. Table 2 shows the calibration curves for the small spoon (top), partial large spoon (middle), and large spoon (bottom). The large spoon, with the largest volume and depth, appears to provide the best curve fit, with an R2 of 0.997. The worst curve fit is for the small spoon, with an R2 of 0.983, while the partially filled large spoon falls in between with an R2 of 0.991. Based on these curve fits, it is recommended to use the large spoon when the quantity of sample is sufficient. When only small sample quantities are available, the partially filled large spoon may be a better option than the small spoon, since the former can be scanned with the 12 mm aperture. 3.4. Continuous flow Only white, un-dyed powder was tested in the case of continuous flow, to illustrate the general feasibility of continuous measurements with this instrument. Two different continuous runs were performed, one using the 12 mm aperture, and one with the 6 mm aperture. The resulting spectra are given in Fig. 7, where each curve indicates a measurement taken at a given elapsed time. From these results, it appears that the spectra for the 12 mm aperture vary over time, while the spectra for the 6 mm aperture are more consistent aside from one outlier time point. The primary cause of variation in continuous measurements is likely to be the surface of the powder bed. Although our setup included a scraping mechanism by the instrument stand, it is difficult to ensure a completely smooth surface. Unlike in static measurements, where the vertical distance is fixed in place by the sample spoon holder, in continuous mode, it is difficult to ensure that the smoothness and correct vertical distance remain after being scraped, as the powder still must travel a short horizontal distance before the measurement is taken. Thus, the 6 mm aperture may be a better option for continuous mode since it captures a smaller area, and therefore will not be as sensitive to macroscopic smoothness issues on the powder surface. However, we recommend that both aperture sizes be tested for any continuous mode setup, and care should be taken to ensure the powder surface is as smooth as possible.
3.3. Sample volume and depth 4. Axial mixing case study Three different sample sizes were tested that varied in volume and depth. The device comes with a small spoon (3.2 mL in volume, 9 mm
The horizontal rotating drum is a commonly used unit operation across industries, including calcination in catalytic manufacturing. For batch operations, the drum or cylinder is placed in a horizontal position and rotated about its axis. For continuous operations, the drum is often arranged in an inclined configuration, typically by a few degrees. Calciners, along with other unit operations, are often characterized by their residence time distribution (RTD) [18]. The RTD can be related to the axial dispersion coefficient using the Taylor dispersion model and Peclet number [19,20]: Peð1−θÞ2 1 EðθÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi e− 4θ 2 πθ=Pe
Fig. 5. Calibration curve for 5 dyeing procedures. The graph shows the percent dyed powder as a function of percent reflectance at a wavelength of 570 nm. The curve was fit with a fourth order polynomial, with an R2 of 0.99975.
where E(θ) is the residence time distribution, θ is the dimensionless time, and Pe is the Peclet number (the ratio of advective to dispersive transport). This approach can be used to characterize continuous systems, and so would be most applicable to pilot-scale systems. To aid in the design and scale-up of calciners, measuring the axial dispersion
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
397
Fig. 6. Calibration curves in the 1–10% dyed powder range, for (a) 3 dyeing procedures, (b) 4 dyeing procedures, (c) 5 dyeing procedures, and (d) 6 dyeing procedures. Fourth order polynomial fits provide R2 values of 0.974, 0.964, 0.984, and 0.968, respectively.
Table 1 Comparison of the effects of aperture size and surface roughness on the calibration curve fits. Calibration curve fit with fourth order polynomial
R2
Aperture size (mm)
Surface (smooth or rough)
12
Smooth
0.997
6
Smooth
0.991
12
Rough
0.869
398
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
Table 2 Comparison of the effects of sample volume and depth on the calibration curve fits. Sample volume (mL)
Sample depth (mm)
Calibration curve fit with fourth order polynomial
R2
3.2
9
0.983
5.4
4
0.991
21.5
14
0.997
coefficient in a bench-scale closed (i.e., non-continuous) system would be advantageous [21]. Using the Peclet number, an axial dispersion coefficient measured in a non-continuous system can be compared to that measured in a continuous flow system. This approach is illustrated using a case study below. It has been found that the axial dispersion of particulate systems can be described using Fick's law [22]:
2
∂c ∂ c ¼ Dax 2 ∂t ∂x
where c is the concentration dyed material, t is the time, x is the axial position, and Dax is the axial dispersion coefficient. In this study, a cylinder 4 cm in diameter and 30 cm in length was loaded with un-dyed FCC in the first 10 cm, dyed FCC in the second 10 cm, and un-dyed FCC in the third 10 cm, to a fill level of 35%. The cylinder was rotated at 25 rpm for 30 min. At 5 min intervals, the cylinder was stopped and 2.5 mL samples were taken every 3 cm. The color spectrophotometer was used to measure the reflectance of each sample (see Fig. 8a), and a calibration curve was used to determine the concentration of dyed material in each sample (see Fig. 8b). Concentration profiles were created as a function of axial position for each time point. Fick's second law was solved
Fig. 7. Spectra for un-dyed powder flowing continuously on a conveyor belt, measured with the (a) 12 mm aperture and (b) 6 mm aperture.
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
399
Fig. 8. (a) Percent reflectance of samples of dyed and un-dyed FCC. (b) Concentration of dyed material in FCC samples calculated using a calibration curve.
numerically, and the axial dispersion coefficient that minimized the residual between the numerical solutions was identified. For three replicates of the six time points, the average value was 0.016 cm2/min with a standard deviation of 0.007 cm2/min. This value is in agreement with axial dispersion coefficients of powders previously found in the literature [21,23]. 5. Conclusions A simple color measurement system that measures reflectance as a function of wavelength was successfully applied to determine tracer powder color concentration. Calibration curves were constructed at a single wavelength to obtain color concentration as a function of reflectance. The measurements were the most sensitive to concentrations less than 10% dyed material, so this was the region used for calibration curve comparisons. All of the system operating conditions were explored to ascertain their effects on calibration curves. Dye strength, aperture size, surface roughness, sample volume and depth, and continuous flow were all tested. Preparing mixtures of varying tracer concentration, static measurements were performed to obtain spectral color information. Five dyeing procedures (5 Sharpies/100 mL powder) was the ideal dye strength for our 80 micron FCC material. The best sample chamber and aperture was the large spoon with the 12 mm aperture. When only a limited amount of material is available, the partial large spoon may be better than the small spoon since the former can be scanned with the larger aperture. A partial large spoon can easily be constructed by the insertion of a flat solid object, such as a piece of cardboard, into the large spoon. With every sample, care should be taken to ensure the sample surface is as smooth as possible. The feasibility of using the spectrophotometer for continuous flow was explored. Both aperture sizes were tested on un-dyed powder flowing continuously on a conveyor belt, with measurements taken every few seconds. Here, the smaller 6 mm aperture may be better because it is likely to be less sensitive to macroscopic surface roughness. Continuous operation will need to be configured for each application, but this setup has the potential to be used as an in-line color tracer measurement method. Finally, we applied our color spectrophotometric technique to an axial dispersion experiment of a mixture of dyed and un-dyed powder in a rotating cylinder. Taking samples across the cylinder over time and analyzing the resulting spectra validated that our method can easily be applied to assess mixing. Although a specific color spectrophotometer was used in these measurements, it is expected that the general recommendations will translate to other colorimeters that measure reflectance as a function of wavelength. For future work, it would be useful to test the particle size limits of the spectrophotometry technique. As one can imagine, as particle size increases, the natural packing fraction decreases, meaning individual particles may start to protrude and produce a more uneven surface. As was shown in the present study, surface roughness causes
non-reproducible calibration curves. It is expected that there is a particle size threshold below which spectrophotometry would be more effective, and above which image analysis may be the better option. Returning to the original motivation for this work, this spectrophotometric technique is ideal for measuring residence time distributions (RTDs). Samples can be collected at the outlet of the unit operation and measured off-line, or the spectrophotometer can be placed directly above the flowing material to take continuous measurements in-line. As it is a simple method that can be applied to both batch and continuous systems, this spectrophotometric measurement technique has the potential to transform the way color concentrations of powder mixtures are quantified. Acknowledgments The authors would like to acknowledge funding from the Rutgers Catalyst Manufacturing Science and Engineering Consortium. The authors would also like to thank Dr. James Scicolone for taking the SEM photographs. References [1] S.L. Lee, T.F. O'Connor, X. Yang, C.N. Cruz, S. Chatterjee, R.D. Madurawe, et al., Modernizing pharmaceutical manufacturing: from batch to continuous production, J. Pharm. Innov. (2015), http://dx.doi.org/10.1007/s12247-015-9215-8. [2] M. Poux, P. Fayolle, J. Bertrand, D. Bridoux, J. Bousquet, Powder mixing: some practical rules applied to agitated systems, Powder Technol. 68 (1991) 213–234, http://dx.doi.org/10.1016/0032-5910(91)80047-M. [3] E.M. Holt, The properties and forming of catalysts and absorbents by granulation, Powder Technol. 140 (2004) 194–202, http://dx.doi.org/10.1016/j.powtec.2004.01. 010. [4] Y. Gao, F.J. Muzzio, M.G. Ierapetritou, A review of the residence time distribution (RTD) applications in solid unit operations, Powder Technol. 228 (2012) 416–423, http://dx.doi.org/10.1016/j.powtec.2012.05.060. [5] S. Tallon, C.E. Davies, B. Barry, Slip velocity and axial dispersion measurements in a gas-solid pipeline using particle tracer analysis, Powder Technol. 99 (1998) 125–131, http://dx.doi.org/10.1016/S0032-5910(98)00095-3. [6] X. Deng, J. Scicolone, X. Han, R.N. Davé, Discrete element method simulation of a conical screen mill: a continuous dry coating device, Chem. Eng. Sci. 125 (2015) 58–74. [7] Y. Gao, B.J. Glasser, M.G. Ierapetritou, A. Cuitino, F.J. Muzzio, J.W. Beeckman, et al., Measurement of residence time distribution in a rotary calciner, AIChE J. 59 (2013) 4068–4076, http://dx.doi.org/10.1002/aic.14175. [8] G. Grasa, J.C. Abanades, A calibration procedure to obtain solid concentrations from digital images of bulk powders, Powder Technol. 114 (2001) 125–128, http://dx.doi. org/10.1016/S0032-5910(00)00262-X. [9] A. Realpe, C. Velázquez, Image processing and analysis for determination of concentrations of powder mixtures, Powder Technol. 134 (2003) 193–200, http://dx.doi. org/10.1016/S0032-5910(03)00138-4. [10] S. Muerza, H. Berthiaux, S. Massol-Chaudeur, G. Thomas, A dynamic study of static mixing using on-line image analysis, Powder Technol. 128 (2002) 195–204, http://dx.doi.org/10.1016/S0032-5910(02)00197-3. [11] H. Berthiaux, V. Mosorov, L. Tomczak, C. Gatumel, J.F. Demeyre, Principal component analysis for characterising homogeneity in powder mixing using image processing techniques, Chem. Eng. Process. Process Intensif. 45 (2006) 397–403, http://dx.doi. org/10.1016/j.cep.2005.10.005. [12] A. Aït Aissa, C. Duchesne, D. Rodrigue, Polymer powders mixing part II: multicomponent mixing dynamics using RGB color analysis, Chem. Eng. Sci. 65 (2010) 3729–3738, http://dx.doi.org/10.1016/j.ces.2010.03.007.
400
H.N. Emady et al. / Powder Technology 286 (2015) 392–400
[13] M. Kheiripour Langroudi, P.R. Mort, G.I. Tardos, Study of powder flow patterns in a Couette cell with axial flow using tracers and solid fraction measurements, Granul. Matter 13 (2011) 541–552, http://dx.doi.org/10.1007/s10035-011-0282-3. [14] T.N. McCaig, Extending the use of visible/near-infrared reflectance spectrophotometers to measure colour of food and agricultural products, Food Res. Int. 35 (2002) 731–736, http://dx.doi.org/10.1016/S0963-9969(02)00068-6. [15] K. Slettengren, P. Heunemann, O. Knuchel, E.J. Windhab, Mixing quality of powder–liquid mixtures studied by near infrared spectroscopy and colorimetry, Powder Technol. 278 (2015) 130–137, http://dx.doi.org/10.1016/j.powtec. 2015.03.020. [16] P. Shenoy, F. Innings, T. Lilliebjelke, C. Jonsson, J. Fitzpatrick, L. Ahrné, Investigation of the application of digital colour imaging to assess the mixture quality of binary food powder mixes, J. Food Eng. 128 (2014) 140–145, http://dx.doi.org/10.1016/j. jfoodeng.2013.12.013. [17] D. Barling, D.A.V. Morton, K. Hapgood, Pharmaceutical dry powder blending and scale-up: maintaining equivalent mixing conditions using a coloured tracer powder, Powder Technol. 270 (2015) 461–469, http://dx.doi.org/10.1016/j.powtec.2014.04. 069.
[18] Y. Gao, A. Vanarase, F. Muzzio, M. Ierapetritou, Characterizing continuous powder mixing using residence time distribution, Chem. Eng. Sci. 66 (2011) 417–425, http://dx.doi.org/10.1016/j.ces.2010.10.045. [19] A.T. Harris, J.F. Davidson, R.B. Thorpe, The influence of the riser exit on the particle residence time distribution in a circulating fluidised bed riser, Chem. Eng. Sci. 58 (2003) 3669–3680, http://dx.doi.org/10.1016/S0009-2509(03)00215-X. [20] A.T. Harris, J.F. Davidson, R.B. Thorpe, Particle residence time distributions in circulating fluidised beds, Chem. Eng. Sci. 58 (2003) 2181–2202, http://dx.doi.org/10. 1016/S0009-2509(03)00082-4. [21] R.G. Sherritt, J. Chaouki, A.K. Mehrotra, L.A. Behie, Axial dispersion in the threedimensional mixing of particles in a rotating drum reactor, Chem. Eng. Sci. 58 (2003) 401–415, http://dx.doi.org/10.1016/S0009-2509(02)00551-1. [22] P.M.C. Lacey, Developments in the theory of particle mixing, J. Appl. Chem. 4 (2007) 257–268, http://dx.doi.org/10.1002/jctb.5010040504. [23] A.S. Bongo Njeng, S. Vitu, M. Clausse, J.-L. Dirion, M. Debacq, Effect of lifter shape and operating parameters on the flow of materials in a pilot rotary kiln: part I. Experimental RTD and axial dispersion study, Powder Technol. 269 (2015) 554–565, http://dx.doi.org/10.1016/j.powtec.2014.03.066.