- Email: [email protected]
Size, shape, and density changes of biomass particles during rapid devolatilization
Fuel 206 (2017) 342–351
Contents lists available at ScienceDirect
Fuel journal homepage: www.elsevier.com/locate/fuel
Full Length Article
Size, shape, and density changes of biomass particles during rapid devolatilization Per Holmgren a,⇑, David R. Wagner a, Anna Strandberg a, Roger Molinder b, Henrik Wiinikka b, Kentaro Umeki c, Markus Broström a a b c
Umeå University, Department of Applied Physics and Electronics, Thermochemical Energy Conversion Laboratory, SE-901 87 Umeå, Sweden RISE Energy Technology Center, Box 726, SE-941 28 Piteå, Sweden Luleå University of Technology, Division of Energy Science, SE-971 87 Luleå, Sweden
h i g h l i g h t s A particle conversion model based on optical particle properties is presented. Shape and density transformations of devolatilizing biomass particles were quantified. Shape descriptors for biomass particles assessed with important differences found. The method and model is suitable for implementing in in-situ applications.
a r t i c l e
i n f o
Article history: Received 17 January 2017 Received in revised form 26 April 2017 Accepted 1 June 2017
Keywords: PIV DTR Pyrolysis Biomass conversion
a b s t r a c t Particle properties such as size, shape and density play significant roles on particle flow and flame propagation in pulverized fuel combustion and gasification. A drop tube furnace allows for experiments at high heating rates similar to those found in large-scale appliances, and was used in this study to carry out experiments on pulverized biomass devolatilization, i.e. detailing the first stage of fuel conversion. The objective of this study was to develop a particle conversion model based on optical information on particle size and shape transformation. Pine stem wood and wheat straw were milled and sieved to three narrow size ranges, rapidly heated in a drop tube setup, and solid residues were characterized using optical methods. Different shape descriptors were evaluated and a shape descriptor based on particle perimeter was found to give significant information for accurate estimation of particle volume. The optical conversion model developed was proven useful and showed good agreement with conversion measured using a reference method based on chemical analysis of non-volatilized ash forming elements. The particle conversion model presented can be implemented as a non-intrusive method for in-situ monitoring of particle conversion, provided density data has been calibrated. Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction Using pulverized biomass in oxygen blown pressurized biomass gasification has proven efficient for obtaining high gas quality, reducing the need for gas cleaning before further upgrading and use. Design and optimization of entrained-flow biomass gasification aims at a high fuel conversion degree while at the same time keeping residence times (reactor size) short. For these reasons, fuel particle size and size distribution are important aspects to consider. Efficient milling of biomass to an acceptable size distribution is one of the bottlenecks in the development of systems based on ⇑ Corresponding author. E-mail address: [email protected] (P. Holmgren). http://dx.doi.org/10.1016/j.fuel.2017.06.009 0016-2361/Ó 2017 Elsevier Ltd. All rights reserved.
entrained-flow gasification since rather small (<1 mm) [1] particles are required, but conversely fine milling is energy demanding and resulting fibrous materials with low bulk densities are associated with feeding problems [2]. Besides the difficulties in fuel preparation and feeding, the size of biomass particles is known to have an effect on the gas and liquid products, where larger particles lead to reduced fuel yields and an increase in tar formation, increasing overall costs [3]. The physical properties of biomass particles such as particle size, morphology and porous structure affect the particle transport and fuel conversion behavior inside entrained flow gasifiers [4]. By determining these morphological changes, namely swelling/ shrinkage, shape and density change behavior during devolatilization, their effect on biomass gasification inside the entrained flow
343
P. Holmgren et al. / Fuel 206 (2017) 342–351
Nomenclature Aparticle Dc Deq D0eq Di D2D,in D2D,out D2D l msample pparticle t
Area of a 2D image of a particle (mm2) Diameter of smallest circumscribed circle outside shape (mm) Sphere equivalent diameter (mm) Sphere diameter of raw fuel (mm) Diameter of largest inscribed circle inside shape (mm) Particle diameter at inlet (mm) Particle diameter at outlet (mm) Circle equivalent diameter (mm) Length (fall height in DTF) (m) Mass of a given sample Perimeter around shape (mm) Residence time (s)
gasifier can be quantified [5–7]. Morphological changes of particles have been studied both during pyrolysis and char gasification, though mainly for coal [7–13]. For small biomass particles (<1.0 mm) and high heating rates (>1000 K s1), few studies that investigated particle morphological changes qualitatively during pyrolysis can be found [7,14–19]. Through the study of fuel conversion and morphological changes, this paper focuses on how reaction parameters affects pyrolysis behavior of two different fuels: pine stem wood and wheat straw. The objective was to develop a detailed particle devolatilization model including information on particle size, shape, density and mass changes. Drop tube furnace experiments were performed with varying residence time (i.e. fall height) and temperature settings, using two biomass fuels with three initial particle (sieve) size classes. Optical information on particle size and shape was used as model input. 2. Material and methods 2.1. Fuel samples Pine stem wood and wheat straw were milled to 0.1–1.0 mm in a Retsch SM 2000 cutting mill after drying at 105 °C. The biomass particles were sieved to various size ranges: 125–150, 400–425 and 600–630 lm. Fuel analysis of the raw fuels is presented in Table 1. The ash content of wheat straw particles differed depending on sieve size, which has also been previously shown by others [20]. These differences are probably due to fractionation of
vin vout
xmin xmax Xds b
c d
qgeo q0geo /Cox /Riley
w
Gas velocity at inlet (m/s) Gas velocity at outlet (m/s) Shortest cord length across shape (mm) Longest cord length across shape (mm) Dry substance conversion degree Diameter evolution coefficient Initial stage density evolution coefficient End stage density evolution coefficient Density based on geometry (kg/m3) Geometric density of raw fuel (kg/m3) Cox circularity Riley circularity Sphericity
different parts of the plant during milling and sieving. High ash content components seem to break into smaller fragments more easily than low ash components. This was not observed for the more homogeneous stem wood. Ash and volatile contents were determined by methods EN 14775 and EN 15148, respectively. Carbon, hydrogen, and nitrogen contents determined by method EN 15104, where oxygen is calculated by difference. Estimated uncertainties for the analysis of inorganic elements can be found in Supplementary Material (S1). 2.2. Drop tube furnace A drop tube furnace (DTF) with an alumina (Al2O3) reactor tube with an inner diameter of 50 mm and a height of 2 m was utilized. The DTF was operated under atmospheric pressure at 900 °C and 1100 °C using five individually controlled heating zones (354 mm long). A syringe feeder [21] was used to supply the biomass at a fuel feeding rate of 5 g h1 and a N2 flow rate of 0.38 nL min1 carrying the fuel particles. Secondary N2 was also supplied from the top of the reactor with a flow rate of rate of 5.0 nL min1, preheated to 350 °C using a Leister heater, and further heated in the first zone before merging with the fuel laden carrier gas. The particle injection tube had an inner diameter of 6 mm and was encased in a water-cooled jacket. Volume fractions of oxygen and carbon monoxide in the product gas were recorded at the exit to validate a leak-free system and stable fuel feeding. Biomass particles were entrained in the N2 flow and were devolatilized as they travelled down the reactor. Particle residence time was varied by changing
Table 1 Fuel analyses for pine stem wood and wheat straw based on dry substance (ds). Composition Ash Content Volatile Matter Fixed Carbon Carbon Hydrogen Nitrogen Oxygen (by difference)
wt%ds wt%ds wt%ds wt%ds wt%ds wt%ds wt%ds
K Na Ca Mg Al Fe Si P S Cl
mg mg mg mg mg mg mg mg mg mg
kg1 kg1 kg1 kg1 kg1 kg1 kg1 kg1 kg1 kg1
Pine
Wheat straw 125–150 mm
Wheat straw 400–425 mm
Wheat straw 600–630 mm
0.4 84.8 14.8 50.6 6.2 0.2 42.6
8.2 73.4 18.4 45.3 5.8 0.7 40
7.1 74.7 18.2 46.0 5.8 0.6 40.5
6.4 75.3 18.3 46.4 5.9 0.6 40.7
813 17 838 229 31 30 183 161 121 26
13500 188 4560 1120 568 419 19400 760 1270 3630
13500 189 3950 936 520 358 14700 680 1180 3540
12100 177 3260 806 431 275 12600 600 1130 3240
344
P. Holmgren et al. / Fuel 206 (2017) 342–351
the fall height inside the reactor. Char particles were quenched using a water-cooled probe inserted from the bottom of the reactor that was moved to correspond to the use of two, three or four heating zones, giving fall heights of 354, 708 and 1062 mm respectively. The particles were collected in cyclones with cut-off diameters of approximately 0.66 mm. A schematic figure of the equipment can be found in Fig. 1. 2.3. Optical equipment Raw fuel and collected char were analyzed using two different commercially available instruments for image analysis: a Camsizer XT (Retsch Technology GmbH) and a LaVision ParticleMaster system (Imager SX 4M CCD camera, LaVision GmbH). The Camsizer XT was used for dynamic image analysis (DIA), to analyze size distribution and shape descriptors. This was done by feeding the sample in an even air stream falling past two high-speed CCD cameras. One camera is optimized for the larger particles in the sample, while the other camera is zoomed, optimized for the smallest fractions. This allows accurate size and shape analysis of the full size range of particles. The LaVision system was used for similar DIA of raw fuel particles, though only using a single CCD camera. It was also configured for static image analysis (SIA), of stationary particles on a flat surface seen from above. DIA was used to measure free fall velocities by taking pictures of particles falling through a separate transparent quartz tube with an inner diameter of 50 mm. DIA of data collected using the LaVision system was done using the software DaVis version 8.1.5. ImageJ v.1.48 [22] was used for density calculations based on SIA data. 2.4. Residence time determination Residence time plays a large role in thermal decomposition and is in the DTF determined by the gas velocity, the slip velocity of a fuel particle and the length of the reactor [16,18]. However residence time could not be measured directly inside the DTF. By use of DIA the particle velocity of the three raw fuel fractions for both fuel types in air was measured and the slip velocity of each fuel was approximated as a function of their measured circle equivalent diameter. The influence of shape was not modelled at this point. The fall height was set to 1000 mm for these measurements.
Within the particle size region of interest, the velocity could be approximated as a second order polynomial function of the particle diameter. These trendline functions were then modified using Stoke’s law to consider the differences in viscosity and density between air at room temperature and hot nitrogen gas. The modified functions were used to calculate the slip velocity of the particles inside the reactor, also taking into account the relevant Reynolds number region (0.4 < Re < 500) based on their measured size [23]. Residence times for the different DTF experiments were calculated as the fall distance divided by the size dependent mean velocities of the particles based on the optically determined entry and exit diameters of the particles (Eq. (1)).
t ¼ 2l=ðf ðD2D;in Þ þ v in þ f ðD2D;out Þ þ v out ðTÞÞ
ð1Þ
Residence time of the particles is given by t, l is the fall length, and f(D) is the empirical function for the slip velocity given by experiments with raw fuel particles and normalized to the given temperature T of the gas in the reactor. D2D,in is the average inlet diameter of the particles based on their observed 2D area, D2D,out is the average outlet diameter of the particles based on their observed 2D area, as described in the section below. vin is the calculated inlet gas velocity and vout in the calculated outlet gas velocity of the gas depending on the temperature in the reactor. 2.5. Quantification of morphology and geometric density The shape, size and density of an irregularly shaped particle in a 2D image can be described in several different ways; the parameters used in this paper are summarized in Table 2. Given known scaling and resolution of the images taken, the area of each pixel was known. By summing the pixels that make up a particle, the geometric 2D area, Aparticle, could be determined and then converted into a circle-equivalent 2D diameter, D2D, (Eq. (2)). Using 2D images to model 3D objects is not entirely straight forward, and in this paper the method described by Bagheri et al. [24] was used to calculate a 3D equivalent volume diameter, Deq, of a particle from 2D images (Eq. (3)). Their findings indicate that diameters of 3D objects based on 2D images can be improved by taking into account particle sphericity, w, where they used Cox circularity, /Cox, for vesicular particles and Riley circularity, /Riley, for
Fig. 1. Schematic figure of the drop tube furnace and the inlet of the quenching sampling probe.
345
P. Holmgren et al. / Fuel 206 (2017) 342–351 Table 2 Equations used to characterize fuel particles. Formula pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D2D ¼ 4Aparticle =p
Notes
Equation (2)
Deq ¼ D2D =1:022w uCox for vesicular particles w uRiley for non-vesicular particles
2D circle-equivalent diameter Equivalent diameter[24] Sphericity
(4)
uCox ¼ 4pAparticle =P2particle pffiffiffiffiffiffiffiffiffiffiffiffiffi uRiley ¼ Di=Dc sample qgeo ¼ Pn m ð4p=3ÞðD Þ3
Cox circularity[25]
(5)
Riley circularity[26]
(6)
Geometric density
(7)
0:29
i¼1
(3)
eq
non-vesicular particles (Eqs. (4–6)). These are equal to 1 for a perfect sphere, and decrease towards zero as the shape becomes less sphere-like. Cox circularity describes how the perimeter of the particle, Pparticle, differs from the perimeter of a circle with the same area as the particle. Riley circularity compares the diameter of the largest inscribed circle, Di, that can fit inside the shape divided by the diameter of the smallest circumscribed circle, DC, that can fit around the outside of the shape. For more information on the equations used see Camsizer documentation, DIN 66141 [3]. Geometric density, qgeo, was determined using Eq. (7), dividing the total mass of all n particles used for SIA image analysis, msample, by the sum of the volumes based on Deq. Char conversion was determined by chemical analysis of ash forming elements, and calculating conversion using refractory elements, e.g. Mg, Ca or Si as a tie component, as described by others [27–29]. For this study Mg was chosen since the high Si content was unevenly distributed in the wheat straw samples, and Mg showed the most stable trends. The method has proven more successful than using the ash content determined by standard methods as tracer, since some components of the ash are likely to be volatilized during the fuel conversion. In this paper, the conversion degree was calculated based on ICP-AES analysis of magnesium content in the dry raw fuel and in the char, and from here on this method will be referred to as the reference method. The conversion estimates had a standard error of 5.7% and 1.4% for pine stem wood and wheat straw respectively. 2.6. Fuel conversion model Optical data were collected at different fall heights of the reactor, size was described by the use of 2D circle-equivalent diameter and sphere equivalent diameter (Eqs. (2) and (3)). Sphere equivalent diameter requires a measurement of particle sphericity (Eq. (4)). In this paper sphericity was assessed through use of Cox circularity and Riley circularity (Eqs (5) and (6)). To calculate geometric density, sphere-equivalent diameter and mass were needed (Eq. (7)). Measured density and size change could be normalized against the raw dry fuel giving relative mass loss, which was used to describe conversion degree. The only input terms needed for the density model after density calibration was size and shape. Negligible agglomeration and fragmentation is assumed. The flowchart from data collection to conversion model is shown in Fig. 2.
Fig. 2. Model design flow scheme.
through the reactor faster, meaning they are subject to shorter residence times as compared to smaller particles given the same fall height. Assessing particle residence time using only theoretical gas residence time gives substantially different values compared to taking slip velocity into account (see Supplementary Material S2). 3.1. Determinations of particle density SIA images of the raw fuel factions were used to measure geometric density (Fig. 3). Interpreting the influence of initial sieve size on geometric density is not entirely straightforward, and the accuracy might relate to the sphericity approximations. Images used for approximating density were taken from above (SIA). The biggest dimensions were on average what was used for size approximations, leading to larger approximated volumes compared to what could be expected if all dimensions were represented equally. This in turn leads to lower calculated densities. However, it does not explain the markedly lower calculated densities for the smallest factions of both fuels observed. In the case of wheat straw, fractionations during sieving could cause size dependent variations in chemical composition, explaining the differences in measured density. In the case of pine stem wood, ash contents were identical for all size fractions; instead higher measured volumes based on geometric shape could be due to fibers protruding from the main bulk of a wood particle. These fibers increase the apparent size of small particles to a higher extent than for larger particles.
3. Results and discussion Gas residence time along the centerline from entry to final sampling point (1062 mm) was calculated using Comsol 4.4. Based on the gas flow and temperature specifications in Section 2.2, it was determined to be 6.4 s. Using Eq. (1), estimated residence times for particles varied between 0.2 and 1.7 s mainly depending on fuel size and fall length, with values close to those reported in a previous study [15]. Large particles have higher slip velocities and fall
Fig. 3. Optically determined geometric densities of raw fuel fractions plotted against their sieve size. Error bars indicate 95% confidence interval.
346
P. Holmgren et al. / Fuel 206 (2017) 342–351
3.2. Particle shape changes Stem wood particles showed a pronounced shape change as they are devolatilized, see Fig. 4(a–c). SEM images indicated that the pine stem wood fuel formed porous plasticized particles. No swelling could be observed for either fuel, as has been reported for biomass particles by others [17,19]. For wheat straw particles the shape change during devolatilization can be seen in Fig. 4(d–f), illustrating transformation from flat flakes to curled sticks. The evolution of particle morphology during conversion in the DTF was assessed in relation to their raw initial state. Particle size and shape for raw fuel fractions are shown in Fig. 5, where both SIA and DIA results are shown. Variations in diameter using SIA and DIA are small, with the smallest fraction of wheat straw being the only sample to give any significant difference. On the other hand, for Cox circularity a systematic difference was found. An explanation for this is that the images taken with the SIA system were seen from above on stationary particles, on average representing particles with the two largest dimensions since non spherical particles tend to lie with the flat side down. DIA takes images from the side of particles tumbling past, leading to a higher representation of the smallest dimension of the particle. Because of this, particles appear rounder when seen from above.
Particle shape was assessed by use of circularity (Eqs. (4) and (5)). Final stem wood char particle circularity was similar for the chars from medium and large size fractions using either Riley or Cox circularity (Fig. 6a and b), both increasing with longer residence time. The smallest size fraction gave substantially lower values using Cox circularity, though the difference decreases with increased residence time. Cox circularity also indicated lower circularity for the raw fuel with decreasing sieve size. This is likely due to Cox circularity being sensitive to the outline of the particles, and therefore to a greater extent affected by small fibers sticking out; smaller particles have higher surface to volume ratio, and hence the impact of surface roughness will be larger on Cox circularity. Measured wheat straw particle circularity increased slightly during initial devolatilization according to either Riley or Cox circularity (Fig. 6c and d). No change in circularity was observed after the initial shape transformation. Another observation was that the smallest size fraction of chars had consistently higher circularity than the medium and large size fractions. These trends were the same independently of circularity definition used. However, Riley circularity gave higher absolute values in all cases, explained by differences in perimeter roughness between the fuels. Riley circularity was found to indicate that medium and large size fractions of stem wood and wheat straw chars are of comparable shape
Fig. 4. SEM image of raw 600–630 mm pine stem wood particles (a), sampled at 354 mm (b) and 1062 mm (c) and of raw 600–630 mm wheat straw particles (d), sampled at 354 mm (e) and 1062 mm fall height (f) at 1100 °C DTF temperature.
P. Holmgren et al. / Fuel 206 (2017) 342–351
347
Fig. 5. Optical diameter of the raw fuels factions (a) and Cox circularity (b), determined with the two optical methods SIA and DIA. Error bars indicate 95% confidence interval.
Fig. 6. Riley circularity for pine stem wood plotted in (a) and Cox circularity in (b). For wheat straw Riley circularity is plotted in (c) and Cox circularity in (d).
(Fig. 6a and c), thus failing to describe the marked geometrical differences found. Using Cox circularity instead, the differences both between pine stem wood and wheat straw, and between the different stem wood size fractions was well described (Fig. 6b and d). Riley circularity was thereby concluded less suitable for describing the shape change of the fuels used. Throughout the rest of this paper an assumption of Cox circularity was used. 3.3. Particle size changes By use of the shape descriptors from the previous section, Eq. (3) can be solved giving a volume equivalent diameter for the particles. In the case of stem wood (Fig. 7a) higher temperature leads to faster shrinkage, while thermal inertia of the larger initial size particles slowed down the rate of shrinkage during the first 0.3 s. Initial size did not seem to play a large role for final size in relative terms. Size appeared to decrease between all measurement points. Taking only size into account, this would indicate that, at least for the larger size fractions, the residence time was probably not long
enough for full devolatilization of stem wood particles. Wheat straw particles behaved differently compared to stem wood (Fig. 7), as it reacted much faster and there was no noticeable size change after the first measuring point for any size fraction, shrinking to 59 ± 6% of initial size. However, the smallest sieve size (125– 150 mm) deviated strongly from the other sizes, shrinking to 82 ± 8% of initial size (95% confidence interval). Given the difference in chemical composition of the smallest wheat straw fraction as compared to the larger ones, the higher amount of ash forming elements could contribute to volume stability during devolatilization, thus decreasing the observed shrinkage. Differences in macromolecular composition, possibly caused by fractionation during milling and sieving, may also affect char formation. Describing how the diameter of a particle changes as a function of particle conversion is often described using Eq. (9), which is a slight variation of an equation proposed by Smith [30] for coal fuel particles, but where the particle equivalent diameter is used instead of the measured circle equivalent diameter. X is the degree of conversion, D0eq is the equivalent diameter of the raw fuel, while
348
P. Holmgren et al. / Fuel 206 (2017) 342–351
Fig. 7. Relative diameter change of particles plotted against residence time, pine stem wood in (a) and wheat straw in (b).
the diameter evolution coefficient b can vary between 0 and ⅓. A value of 0 describes a particle where mass loss occurs strictly through density loss while the diameter remains constant, and a value of ⅓ describes a situation where the density remains constant and mass is lost solely through shrinkage. To quantify the degree of conversion density is needed.
Deq ¼ D0eq ð1 XÞb
b¼0
Density loss model
b ¼ 1=3 Shrinking particle model
ð9Þ
conversion proceeds. The term q0geo is the geometric density of the raw fuel while the density evolution coefficients c and d fulfills the same role as b in Eq. (9), though not limited to a maximum value of ⅓. One limitation with this equation is that density at full mass conversion (X = 1) would be equal to raw fuel density. However this model is developed for use on a dry fuel basis, and at X = 1 the particle will have a volume of zero (see Eq. (9)). The char residue and ash composition of observable particles will prevent X from reaching 1.
qgeo ¼ q0geo ðð1 XÞc þ X d Þ 3.4. Particle density changes The weight of the raw fuel particles and char particles was carefully measured and then analyzed by SIA to acquire their volume equivalent diameter. This was translated to total sample volume, and thus average particle density was determined (Eq. (7)). Because of experimental difficulties in accurately weighing and collecting useful images for the medium and smallest fuel fractions, only the 600–630 mm particles were used for density change calculations. Pine stem wood particles experienced an initial decrease in optically determined geometric density as they were devolatilized, followed by an increase in density (Fig. 8a), likely due to the particle undergoing plasticization with subsequent shrinking of pores and internal cavities, compare with Fig. 4(c). Higher temperatures led to a slightly higher initial density drop. This could be explained by higher heating rates leading to a more violent release of volatiles. Density change for wheat straw was initially slow followed by a steep decrease somewhere between 0.3 s and 0.6 s residence time, after which density was stabilized (Fig. 8b), indicating that devolatilization was completed. In this paper, modeling density change during devolatilization is done using a similar approach as for diameter change, though the equation contains one additional term containing X (conversion), see Eq. (10). This allows the density to move from density loss to shrinking particle behavior (or the other way around) as
ð10Þ
Eq. (10) can be extended to cover full conversion to pure ash by the inclusion of a constant before the second X term in the equation, describing the ratio between the density of a pure ash particle and that of a raw fuel particle. 3.5. Fuel conversion model For the large sized fuel fractions geometric density and volume equivalent size were directly measured. By combining relative volume change and relative density change, conversion degree could be assessed and so b, c and d in Eqs. (9) and (10) could be determined for the different fuels. See Table 3 for details. By multiplying the relative volume change with the relative density change of particles, mass loss on a dry basis was assessed by use of an optical conversion model (OCM) according to Eq. (11). Table 3 Empirical constants used to model conversion. Detailed fitting results can be found in Supplementary Material (S3 And S4). Biomass type
Temperature (°C)
b
c
d
Pine stem wood Pine stem wood Wheat straw Wheat straw
900 1100 900 1100
0.33 0.31 0.19 0.19
0.66 0.47 1.48 1.11
1.60 4.15 0.90 3.69
Fig. 8. The relative density change of pine stem wood (a) and wheat straw (b) particles as a function of residence time. Error bars represent 95% confidence interval.
P. Holmgren et al. / Fuel 206 (2017) 342–351
X ¼1
Deq D0eq
0
3 bc 1 Deq b @ Deq þ 1 D0eq D0eq
!d 1 A
ð11Þ
A particle that reaches a size and density change that corresponds to 100% mass conversion was defined as fully converted. The relative densities of the medium and small size fractions could then be estimated based on their relative size, and thus mass conversion could be predicted. The flow scheme is illustrated in Fig. 9. This mass loss was compared with fuel conversion estimated using the reference method, see Fig. 10. Error bars correspond to one standard deviation. 3.5.1. Conversion of pine stem wood particles At 900 °C the OCM gave a slightly but significantly sharper conversion curve for smaller initial particle size, which was expected due to lower thermal inertia and thus higher internal heating rates
Fig. 9. Modified model design flow scheme due to limited density data.
349
for smaller particles. The reference method, on the other hand, indicated no dependence on initial size on conversion as all samples seem to follow a similar curve. One sample analyzed using the reference method had a negative mass loss in Fig. 10(a), likely due to a weighing error during analysis. This illustrates the sensitivity to experimental errors not covered within the analytical uncertainties, and may explain why no size dependence on initial conversion was found. The two methods can also be compared at 1100 °C in Fig. 10(b) where the size dependence is more obvious. Details are available in Supplementary Material (S5). After residence times of approximately 0.25 s the largest stem wood fuel fraction (600–630 mm sieve size) was converted by 30% running the reactor at 1100 °C, while the medium size was converted to around 70%. One standard deviation overlap between the two methods in all cases indicates comparable accuracies.
3.5.2. Conversion of wheat straw particles For wheat straw the differences were consistent at both temperatures, though larger at 900 °C. 600–630 mm sized wheat straw particles gave higher conversion values using OCM compared to the reference values, and lower values for the 125–150 mm fraction (Fig. 10c and Fig. 10d). 400–425 mm particles showed no difference between the methods used. The low conversion estimates of the smallest wheat straw faction at 900 °C was expected due to the way the shrinkage behavior of the smallest wheat straw particles deviated from the other size fractions (see Fig. 7b), however the smallest faction at 1100 °C gave conversion estimates that were in close agreement with the reference data. Details are available in Supplementary Material (S6). According to the optical conversion estimations the initial size of the wheat straw particles did not seem to have an impact on conversion rates within the size range tested. This could be explained by the particle shapes where thickness is the limiting factor for heat transfer, and in this case it is far smaller than the measurements characterizing the particle size. Small differences could be seen both for initial size and the temperature used using the reference method, even though data are a bit scattered.
Fig. 10. Particle conversion plotted against residence time for pine stem wood at 900 °C (a), pine stem wood at 1100 °C (b), wheat straw at 900 °C (c) and wheat straw at 1100 °C.
350
P. Holmgren et al. / Fuel 206 (2017) 342–351
3.5.3. Comparing wood and straw conversion models Considering the much higher ash content of wheat straw, it was unproblematic to gather enough char and thus ash for accurate ICP-AES analysis, whereas some of the stem wood experiments did not yield enough char for successful analysis to be performed. Another aspect on the reference method, using tracking of nonvolatilized ash forming elements, is that it relies on the element not being lost from the fuel particle during conversion. If fragments are formed and the element is not evenly dispersed throughout the particle, non-representative parts may be lost, causing errors in the estimations. This could be the case for the wheat straw particles since they are rather heterogeneous in composition, coming from different parts of the plant, potentially explaining the somewhat scattered trends seen for conversion estimated by the reference method. However, the none-destructive optical method worked comparable or better than the ICP-AES reference method for stem wood, probably due to stem wood particles being well described by the shape factors used. For directly analyzing conversion degree of solid fuel particles this gives an alternative to time-consuming and costly lab analysis of low-ash fuels. A reactor with optical ports would also enable in-situ measurements of conversion at different stages of a reactor. Considering how size and density change in different ways for the different biofuels tested, it is also clear that conversion simulation models for these kinds of fuels need to use different assumptions on how mass loss occurs depending on biomass. For stem wood, a shrinking particle model might work well for describing the size change, though the changes in density seem to be related to the formation of large pores and cavities during the initial rapid release of volatiles. For wheat straw, on the other hand, conversion seemed to occur in two steps; first through shrinking, and then as density loss. The wheat straw particles never approached a shape that is easily quantified by traditional shape descriptors such as aspect ratio and diameter, and shape aspects therefore requires further studies.
Combining volume and density information, the conversion model was formulated. The modeled conversions were compared with those estimated by a method based on chemical analysis of samples, and a mass balance based on assumptions of some of the ash forming elements not being volatilized during conversion. In the case of pine stem wood, the conversion estimates were in close agreement with values provided by the reference method. Besides accurate description of the transformation, it gave information indicating that thermal inertia limitations limited conversion of the largest size fraction used. Describing volume and density change of wheat straw particles was less accurate, mainly due to difficulties in modeling size and shape change of the very irregularly shaped particles. A strong size dependence in shrinking behavior was found, possibly due to differences in chemical composition originating from the milling and sieving. The optical conversion model developed can be used for multiphase flow modeling purposes or for in-situ monitoring of particle conversion in thermochemical reactors if optical access is provided. For predictions on mass basis, accurate estimations of density is needed, e.g. as described in this paper. Acknowledgement The authors would like to thank the Swedish Energy Agency through the Swedish Centre for Biomass Gasification (SFC), and the Kempe foundation for financial support. The authors would also like to thank Bio4Energy (B4E), a strategic research environment appointed by the Swedish government, for supporting this work. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.fuel.2017.06.009. References
4. Conclusions Size and shape changes of biomass particles during high heating rate devolatilization were studied in the pyrolysis regime using a drop tube furnace, extractive sampling and image analysis. Besides detailed transformation descriptions of the two fuels studied (pine wood and wheat straw powders), a conversion model was developed based on optical 2D data. Accurate models require useful shape descriptors to give simplified but yet correct representations of the shapes analyzed. The two fuels studied had very different initial particle properties, and the differences were even more pronounced during conversion where the pine wood particles formed plasticized hollow spheres, whereas the flake-like wheat straw particles curled and did not shrink as much. The model developed in this paper uses a particle diameter based on 2D particle area corrected by use of circularity based on measured perimeter of a particle. For this purpose, common shape descriptors were evaluated and Cox circularity (the ratio between the perimeter of a circle and the real shape perimeter, assuming equal shape area) was found best suited for describing the shape change of irregularly shaped biomass and char particles. Besides the volumetric change, a mass based conversion requires information on particle density. The density was estimated by weighting and approximating the volume of representative particles. From the density trends, it was found that the density of pine stem wood particles remain rather constant during conversion, whereas wheat straw char had significantly lower density comparted to the original fuel.
[1] Ahmad AA, Zawawi NA, Kasim FH, Inayat A, Khasri A. Assessing the gasification performance of biomass: A review on biomass gasification process conditions, optimization and economic evaluation. Renew Sust Energ Rev 2016;53:1333–47. [2] Dai JJ, Cui HP, Grace JR. Biomass feeding for thermochemical reactors. Prog Energy Combust 2012;38:716–36. [3] Li SG, Xu SP, Liu SQ, Yang C, Lu QH. Fast pyrolysis of biomass in free-fall reactor for hydrogen-rich gas. Fuel Process Technol 2004;85:1201–11. [4] Dupont C, Boissonnet G, Seiler J-M, Gauthier P, Schweich D. Study about the kinetic processes of biomass steam gasification. Fuel 2007;86:32–40. [5] Lu H, Ip E, Scott J, Foster P, Vickers M, Baxter LL. Effects of particle shape and size on devolatilization of biomass particle. Fuel 2010;89:1156–68. [6] Davidsson KO, Pettersson JBC. Birch wood particle shrinkage during rapid pyrolysis. Fuel 2002;81:263–70. [7] Tolvanen H, Kokko L, Raiko R. Fast pyrolysis of coal, peat, and torrefied wood: Mass loss study with a drop-tube reactor, particle geometry analysis, and kinetics modeling. Fuel 2013;111:148–56. [8] Saastamoinen J, Aho M, Moilanen A, Sorensen LH, Clausen S, Berg M. Burnout of pulverized biomass particles in large scale boiler - Single particle model approach. Biomass Bioenergy 2010;34:728–36. [9] Senneca O, Urciuolo M, Chirone R, Cumbo D. An experimental study of fragmentation of coals during fast pyrolysis at high temperature and pressure. Fuel 2011;90:2931–8. [10] Yu JL, Strezov V, Lucas J, Wall T. Swelling behaviour of individual coal particles in the single particle reactor. Fuel 2003;82:1977–87. [11] Zhang LA, Binner E, Chen LG, Qiao Y, Li CZ, Bhattacharya S, et al. Experimental investigation of the combustion of Bituminous coal in air and O-2/CO2 mixtures: 1 particle imaging of the combustion of coal and char. Energy Fuels 2010;24:4803–11. [12] Hodge EM, Roberts DG, Harris DJ, Stubington JF. The Significance of Char Morphology to the Analysis of High-Temperature Char-CO2 Reaction Rates. Energy Fuels 2010;24:100–7. [13] Costa FF, Costa M. Particle fragmentation of raw and torrefied biomass during combustion in a drop tube furnace. Fuel 2015;159:530–7. [14] Zhang Y, Kajitani S, Ashizawa M, Miura K. Peculiarities of rapid pyrolysis of biomass covering medium- and high-temperature ranges. Energy Fuels 2006;20:2705–12.
P. Holmgren et al. / Fuel 206 (2017) 342–351 [15] Dupont C, Chen L, Cances J, Commandre JM, Cuoci A, Pierucci S, et al. Biomass pyrolysis: Kinetic modelling and experimental validation under high temperature and flash heating rate conditions. J Anal Appl Pyrol 2009;85:260–7. [16] Commandre JM, Lahmidi H, Salvador S, Dupassieux N. Pyrolysis of wood at high temperature: The influence of experimental parameters on gaseous products. Fuel Process Technol 2011;92:837–44. [17] Jarvis MW, Haas TJ, Donohoe BS, Daily JW, Gaston KR, Frederick WJ, et al. Elucidation of biomass pyrolysis products using a laminar entrained flow reactor and char particle imaging. Energy Fuels 2011;25:324–36. [18] Dupont C, Commandre JM, Gauthier P, Boissonnet G, Salvador S, Schweich D, et al. Biomass pyrolysis experiments in an analytical entrained flow reactor between 1073 K and 1273 K. Fuel 2008;87(1073):1155–64. [19] Umeki K, Kirtania K, Chen LG, Bhattacharya S. Fuel particle conversion of pulverized biomass during pyrolysis in an entrained flow reactor. Ind Eng Chem Res 2012;51:13973–9. [20] Tamaki Y, Mazza G. Measurement of structural carbohydrates, lignins, and micro-components of straw and shives: Effects of extractives, particle size and crop species. Ind Crop Prod 2010;31:534–41. [21] Molinder R, Wiinikka H. Feeding small biomass particles at low rates. Powder Technol 2015;269:240–6.
351
[22] ImageJ (Software), in, http://imagej.nih.gov/ij/. [23] Bonadonna GGJEC, Sparks RSJ. Thickness variations and volume estimates of tephra fall deposits. J Volcanol Geoth Res 1998;81. [24] Bagheri GH, Bonadonna C, Manzella I, Vonlanthen P. On the characterization of size and shape of irregular particles. Powder Technol 2015;270:141–53. [25] Cox EP. A method of assigning numerical and percentage values to the degree of roundness of sand grains. J Paleontol 1927;179–183. [26] Riley NA. Projection sphericity. J Sediment Res 1941;94–95. [27] Borrego AG, Alvarez D. Comparison of chars obtained under oxy-fuel and conventional pulverized coal combustion atmospheres. Energy Fuels 2007;21:3171–9. [28] Shah KV, Cieplik MK, Betrand CI, van de Kamp WL, Vuthaluru HB. A kineticempirical model for particle size distribution evolution during pulverised fuel combustion. Fuel 2010;89:2438–47. [29] Li S. Char-Slag Transition During Pulverized Coal Gasification Ph.D Thesis. Department of Chemical Engineering, The University of Utah; 2010. p. 73. [30] Smith IW, The Combustion Rates of Coal Chars: A Review, Nineteenth Symposium (International) on Combustion/The Combustion Institute, (1982).
Contents lists available at ScienceDirect
Fuel journal homepage: www.elsevier.com/locate/fuel
Full Length Article
Size, shape, and density changes of biomass particles during rapid devolatilization Per Holmgren a,⇑, David R. Wagner a, Anna Strandberg a, Roger Molinder b, Henrik Wiinikka b, Kentaro Umeki c, Markus Broström a a b c
Umeå University, Department of Applied Physics and Electronics, Thermochemical Energy Conversion Laboratory, SE-901 87 Umeå, Sweden RISE Energy Technology Center, Box 726, SE-941 28 Piteå, Sweden Luleå University of Technology, Division of Energy Science, SE-971 87 Luleå, Sweden
h i g h l i g h t s A particle conversion model based on optical particle properties is presented. Shape and density transformations of devolatilizing biomass particles were quantified. Shape descriptors for biomass particles assessed with important differences found. The method and model is suitable for implementing in in-situ applications.
a r t i c l e
i n f o
Article history: Received 17 January 2017 Received in revised form 26 April 2017 Accepted 1 June 2017
Keywords: PIV DTR Pyrolysis Biomass conversion
a b s t r a c t Particle properties such as size, shape and density play significant roles on particle flow and flame propagation in pulverized fuel combustion and gasification. A drop tube furnace allows for experiments at high heating rates similar to those found in large-scale appliances, and was used in this study to carry out experiments on pulverized biomass devolatilization, i.e. detailing the first stage of fuel conversion. The objective of this study was to develop a particle conversion model based on optical information on particle size and shape transformation. Pine stem wood and wheat straw were milled and sieved to three narrow size ranges, rapidly heated in a drop tube setup, and solid residues were characterized using optical methods. Different shape descriptors were evaluated and a shape descriptor based on particle perimeter was found to give significant information for accurate estimation of particle volume. The optical conversion model developed was proven useful and showed good agreement with conversion measured using a reference method based on chemical analysis of non-volatilized ash forming elements. The particle conversion model presented can be implemented as a non-intrusive method for in-situ monitoring of particle conversion, provided density data has been calibrated. Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction Using pulverized biomass in oxygen blown pressurized biomass gasification has proven efficient for obtaining high gas quality, reducing the need for gas cleaning before further upgrading and use. Design and optimization of entrained-flow biomass gasification aims at a high fuel conversion degree while at the same time keeping residence times (reactor size) short. For these reasons, fuel particle size and size distribution are important aspects to consider. Efficient milling of biomass to an acceptable size distribution is one of the bottlenecks in the development of systems based on ⇑ Corresponding author. E-mail address: [email protected] (P. Holmgren). http://dx.doi.org/10.1016/j.fuel.2017.06.009 0016-2361/Ó 2017 Elsevier Ltd. All rights reserved.
entrained-flow gasification since rather small (<1 mm) [1] particles are required, but conversely fine milling is energy demanding and resulting fibrous materials with low bulk densities are associated with feeding problems [2]. Besides the difficulties in fuel preparation and feeding, the size of biomass particles is known to have an effect on the gas and liquid products, where larger particles lead to reduced fuel yields and an increase in tar formation, increasing overall costs [3]. The physical properties of biomass particles such as particle size, morphology and porous structure affect the particle transport and fuel conversion behavior inside entrained flow gasifiers [4]. By determining these morphological changes, namely swelling/ shrinkage, shape and density change behavior during devolatilization, their effect on biomass gasification inside the entrained flow
343
P. Holmgren et al. / Fuel 206 (2017) 342–351
Nomenclature Aparticle Dc Deq D0eq Di D2D,in D2D,out D2D l msample pparticle t
Area of a 2D image of a particle (mm2) Diameter of smallest circumscribed circle outside shape (mm) Sphere equivalent diameter (mm) Sphere diameter of raw fuel (mm) Diameter of largest inscribed circle inside shape (mm) Particle diameter at inlet (mm) Particle diameter at outlet (mm) Circle equivalent diameter (mm) Length (fall height in DTF) (m) Mass of a given sample Perimeter around shape (mm) Residence time (s)
gasifier can be quantified [5–7]. Morphological changes of particles have been studied both during pyrolysis and char gasification, though mainly for coal [7–13]. For small biomass particles (<1.0 mm) and high heating rates (>1000 K s1), few studies that investigated particle morphological changes qualitatively during pyrolysis can be found [7,14–19]. Through the study of fuel conversion and morphological changes, this paper focuses on how reaction parameters affects pyrolysis behavior of two different fuels: pine stem wood and wheat straw. The objective was to develop a detailed particle devolatilization model including information on particle size, shape, density and mass changes. Drop tube furnace experiments were performed with varying residence time (i.e. fall height) and temperature settings, using two biomass fuels with three initial particle (sieve) size classes. Optical information on particle size and shape was used as model input. 2. Material and methods 2.1. Fuel samples Pine stem wood and wheat straw were milled to 0.1–1.0 mm in a Retsch SM 2000 cutting mill after drying at 105 °C. The biomass particles were sieved to various size ranges: 125–150, 400–425 and 600–630 lm. Fuel analysis of the raw fuels is presented in Table 1. The ash content of wheat straw particles differed depending on sieve size, which has also been previously shown by others [20]. These differences are probably due to fractionation of
vin vout
xmin xmax Xds b
c d
qgeo q0geo /Cox /Riley
w
Gas velocity at inlet (m/s) Gas velocity at outlet (m/s) Shortest cord length across shape (mm) Longest cord length across shape (mm) Dry substance conversion degree Diameter evolution coefficient Initial stage density evolution coefficient End stage density evolution coefficient Density based on geometry (kg/m3) Geometric density of raw fuel (kg/m3) Cox circularity Riley circularity Sphericity
different parts of the plant during milling and sieving. High ash content components seem to break into smaller fragments more easily than low ash components. This was not observed for the more homogeneous stem wood. Ash and volatile contents were determined by methods EN 14775 and EN 15148, respectively. Carbon, hydrogen, and nitrogen contents determined by method EN 15104, where oxygen is calculated by difference. Estimated uncertainties for the analysis of inorganic elements can be found in Supplementary Material (S1). 2.2. Drop tube furnace A drop tube furnace (DTF) with an alumina (Al2O3) reactor tube with an inner diameter of 50 mm and a height of 2 m was utilized. The DTF was operated under atmospheric pressure at 900 °C and 1100 °C using five individually controlled heating zones (354 mm long). A syringe feeder [21] was used to supply the biomass at a fuel feeding rate of 5 g h1 and a N2 flow rate of 0.38 nL min1 carrying the fuel particles. Secondary N2 was also supplied from the top of the reactor with a flow rate of rate of 5.0 nL min1, preheated to 350 °C using a Leister heater, and further heated in the first zone before merging with the fuel laden carrier gas. The particle injection tube had an inner diameter of 6 mm and was encased in a water-cooled jacket. Volume fractions of oxygen and carbon monoxide in the product gas were recorded at the exit to validate a leak-free system and stable fuel feeding. Biomass particles were entrained in the N2 flow and were devolatilized as they travelled down the reactor. Particle residence time was varied by changing
Table 1 Fuel analyses for pine stem wood and wheat straw based on dry substance (ds). Composition Ash Content Volatile Matter Fixed Carbon Carbon Hydrogen Nitrogen Oxygen (by difference)
wt%ds wt%ds wt%ds wt%ds wt%ds wt%ds wt%ds
K Na Ca Mg Al Fe Si P S Cl
mg mg mg mg mg mg mg mg mg mg
kg1 kg1 kg1 kg1 kg1 kg1 kg1 kg1 kg1 kg1
Pine
Wheat straw 125–150 mm
Wheat straw 400–425 mm
Wheat straw 600–630 mm
0.4 84.8 14.8 50.6 6.2 0.2 42.6
8.2 73.4 18.4 45.3 5.8 0.7 40
7.1 74.7 18.2 46.0 5.8 0.6 40.5
6.4 75.3 18.3 46.4 5.9 0.6 40.7
813 17 838 229 31 30 183 161 121 26
13500 188 4560 1120 568 419 19400 760 1270 3630
13500 189 3950 936 520 358 14700 680 1180 3540
12100 177 3260 806 431 275 12600 600 1130 3240
344
P. Holmgren et al. / Fuel 206 (2017) 342–351
the fall height inside the reactor. Char particles were quenched using a water-cooled probe inserted from the bottom of the reactor that was moved to correspond to the use of two, three or four heating zones, giving fall heights of 354, 708 and 1062 mm respectively. The particles were collected in cyclones with cut-off diameters of approximately 0.66 mm. A schematic figure of the equipment can be found in Fig. 1. 2.3. Optical equipment Raw fuel and collected char were analyzed using two different commercially available instruments for image analysis: a Camsizer XT (Retsch Technology GmbH) and a LaVision ParticleMaster system (Imager SX 4M CCD camera, LaVision GmbH). The Camsizer XT was used for dynamic image analysis (DIA), to analyze size distribution and shape descriptors. This was done by feeding the sample in an even air stream falling past two high-speed CCD cameras. One camera is optimized for the larger particles in the sample, while the other camera is zoomed, optimized for the smallest fractions. This allows accurate size and shape analysis of the full size range of particles. The LaVision system was used for similar DIA of raw fuel particles, though only using a single CCD camera. It was also configured for static image analysis (SIA), of stationary particles on a flat surface seen from above. DIA was used to measure free fall velocities by taking pictures of particles falling through a separate transparent quartz tube with an inner diameter of 50 mm. DIA of data collected using the LaVision system was done using the software DaVis version 8.1.5. ImageJ v.1.48 [22] was used for density calculations based on SIA data. 2.4. Residence time determination Residence time plays a large role in thermal decomposition and is in the DTF determined by the gas velocity, the slip velocity of a fuel particle and the length of the reactor [16,18]. However residence time could not be measured directly inside the DTF. By use of DIA the particle velocity of the three raw fuel fractions for both fuel types in air was measured and the slip velocity of each fuel was approximated as a function of their measured circle equivalent diameter. The influence of shape was not modelled at this point. The fall height was set to 1000 mm for these measurements.
Within the particle size region of interest, the velocity could be approximated as a second order polynomial function of the particle diameter. These trendline functions were then modified using Stoke’s law to consider the differences in viscosity and density between air at room temperature and hot nitrogen gas. The modified functions were used to calculate the slip velocity of the particles inside the reactor, also taking into account the relevant Reynolds number region (0.4 < Re < 500) based on their measured size [23]. Residence times for the different DTF experiments were calculated as the fall distance divided by the size dependent mean velocities of the particles based on the optically determined entry and exit diameters of the particles (Eq. (1)).
t ¼ 2l=ðf ðD2D;in Þ þ v in þ f ðD2D;out Þ þ v out ðTÞÞ
ð1Þ
Residence time of the particles is given by t, l is the fall length, and f(D) is the empirical function for the slip velocity given by experiments with raw fuel particles and normalized to the given temperature T of the gas in the reactor. D2D,in is the average inlet diameter of the particles based on their observed 2D area, D2D,out is the average outlet diameter of the particles based on their observed 2D area, as described in the section below. vin is the calculated inlet gas velocity and vout in the calculated outlet gas velocity of the gas depending on the temperature in the reactor. 2.5. Quantification of morphology and geometric density The shape, size and density of an irregularly shaped particle in a 2D image can be described in several different ways; the parameters used in this paper are summarized in Table 2. Given known scaling and resolution of the images taken, the area of each pixel was known. By summing the pixels that make up a particle, the geometric 2D area, Aparticle, could be determined and then converted into a circle-equivalent 2D diameter, D2D, (Eq. (2)). Using 2D images to model 3D objects is not entirely straight forward, and in this paper the method described by Bagheri et al. [24] was used to calculate a 3D equivalent volume diameter, Deq, of a particle from 2D images (Eq. (3)). Their findings indicate that diameters of 3D objects based on 2D images can be improved by taking into account particle sphericity, w, where they used Cox circularity, /Cox, for vesicular particles and Riley circularity, /Riley, for
Fig. 1. Schematic figure of the drop tube furnace and the inlet of the quenching sampling probe.
345
P. Holmgren et al. / Fuel 206 (2017) 342–351 Table 2 Equations used to characterize fuel particles. Formula pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D2D ¼ 4Aparticle =p
Notes
Equation (2)
Deq ¼ D2D =1:022w uCox for vesicular particles w uRiley for non-vesicular particles
2D circle-equivalent diameter Equivalent diameter[24] Sphericity
(4)
uCox ¼ 4pAparticle =P2particle pffiffiffiffiffiffiffiffiffiffiffiffiffi uRiley ¼ Di=Dc sample qgeo ¼ Pn m ð4p=3ÞðD Þ3
Cox circularity[25]
(5)
Riley circularity[26]
(6)
Geometric density
(7)
0:29
i¼1
(3)
eq
non-vesicular particles (Eqs. (4–6)). These are equal to 1 for a perfect sphere, and decrease towards zero as the shape becomes less sphere-like. Cox circularity describes how the perimeter of the particle, Pparticle, differs from the perimeter of a circle with the same area as the particle. Riley circularity compares the diameter of the largest inscribed circle, Di, that can fit inside the shape divided by the diameter of the smallest circumscribed circle, DC, that can fit around the outside of the shape. For more information on the equations used see Camsizer documentation, DIN 66141 [3]. Geometric density, qgeo, was determined using Eq. (7), dividing the total mass of all n particles used for SIA image analysis, msample, by the sum of the volumes based on Deq. Char conversion was determined by chemical analysis of ash forming elements, and calculating conversion using refractory elements, e.g. Mg, Ca or Si as a tie component, as described by others [27–29]. For this study Mg was chosen since the high Si content was unevenly distributed in the wheat straw samples, and Mg showed the most stable trends. The method has proven more successful than using the ash content determined by standard methods as tracer, since some components of the ash are likely to be volatilized during the fuel conversion. In this paper, the conversion degree was calculated based on ICP-AES analysis of magnesium content in the dry raw fuel and in the char, and from here on this method will be referred to as the reference method. The conversion estimates had a standard error of 5.7% and 1.4% for pine stem wood and wheat straw respectively. 2.6. Fuel conversion model Optical data were collected at different fall heights of the reactor, size was described by the use of 2D circle-equivalent diameter and sphere equivalent diameter (Eqs. (2) and (3)). Sphere equivalent diameter requires a measurement of particle sphericity (Eq. (4)). In this paper sphericity was assessed through use of Cox circularity and Riley circularity (Eqs (5) and (6)). To calculate geometric density, sphere-equivalent diameter and mass were needed (Eq. (7)). Measured density and size change could be normalized against the raw dry fuel giving relative mass loss, which was used to describe conversion degree. The only input terms needed for the density model after density calibration was size and shape. Negligible agglomeration and fragmentation is assumed. The flowchart from data collection to conversion model is shown in Fig. 2.
Fig. 2. Model design flow scheme.
through the reactor faster, meaning they are subject to shorter residence times as compared to smaller particles given the same fall height. Assessing particle residence time using only theoretical gas residence time gives substantially different values compared to taking slip velocity into account (see Supplementary Material S2). 3.1. Determinations of particle density SIA images of the raw fuel factions were used to measure geometric density (Fig. 3). Interpreting the influence of initial sieve size on geometric density is not entirely straightforward, and the accuracy might relate to the sphericity approximations. Images used for approximating density were taken from above (SIA). The biggest dimensions were on average what was used for size approximations, leading to larger approximated volumes compared to what could be expected if all dimensions were represented equally. This in turn leads to lower calculated densities. However, it does not explain the markedly lower calculated densities for the smallest factions of both fuels observed. In the case of wheat straw, fractionations during sieving could cause size dependent variations in chemical composition, explaining the differences in measured density. In the case of pine stem wood, ash contents were identical for all size fractions; instead higher measured volumes based on geometric shape could be due to fibers protruding from the main bulk of a wood particle. These fibers increase the apparent size of small particles to a higher extent than for larger particles.
3. Results and discussion Gas residence time along the centerline from entry to final sampling point (1062 mm) was calculated using Comsol 4.4. Based on the gas flow and temperature specifications in Section 2.2, it was determined to be 6.4 s. Using Eq. (1), estimated residence times for particles varied between 0.2 and 1.7 s mainly depending on fuel size and fall length, with values close to those reported in a previous study [15]. Large particles have higher slip velocities and fall
Fig. 3. Optically determined geometric densities of raw fuel fractions plotted against their sieve size. Error bars indicate 95% confidence interval.
346
P. Holmgren et al. / Fuel 206 (2017) 342–351
3.2. Particle shape changes Stem wood particles showed a pronounced shape change as they are devolatilized, see Fig. 4(a–c). SEM images indicated that the pine stem wood fuel formed porous plasticized particles. No swelling could be observed for either fuel, as has been reported for biomass particles by others [17,19]. For wheat straw particles the shape change during devolatilization can be seen in Fig. 4(d–f), illustrating transformation from flat flakes to curled sticks. The evolution of particle morphology during conversion in the DTF was assessed in relation to their raw initial state. Particle size and shape for raw fuel fractions are shown in Fig. 5, where both SIA and DIA results are shown. Variations in diameter using SIA and DIA are small, with the smallest fraction of wheat straw being the only sample to give any significant difference. On the other hand, for Cox circularity a systematic difference was found. An explanation for this is that the images taken with the SIA system were seen from above on stationary particles, on average representing particles with the two largest dimensions since non spherical particles tend to lie with the flat side down. DIA takes images from the side of particles tumbling past, leading to a higher representation of the smallest dimension of the particle. Because of this, particles appear rounder when seen from above.
Particle shape was assessed by use of circularity (Eqs. (4) and (5)). Final stem wood char particle circularity was similar for the chars from medium and large size fractions using either Riley or Cox circularity (Fig. 6a and b), both increasing with longer residence time. The smallest size fraction gave substantially lower values using Cox circularity, though the difference decreases with increased residence time. Cox circularity also indicated lower circularity for the raw fuel with decreasing sieve size. This is likely due to Cox circularity being sensitive to the outline of the particles, and therefore to a greater extent affected by small fibers sticking out; smaller particles have higher surface to volume ratio, and hence the impact of surface roughness will be larger on Cox circularity. Measured wheat straw particle circularity increased slightly during initial devolatilization according to either Riley or Cox circularity (Fig. 6c and d). No change in circularity was observed after the initial shape transformation. Another observation was that the smallest size fraction of chars had consistently higher circularity than the medium and large size fractions. These trends were the same independently of circularity definition used. However, Riley circularity gave higher absolute values in all cases, explained by differences in perimeter roughness between the fuels. Riley circularity was found to indicate that medium and large size fractions of stem wood and wheat straw chars are of comparable shape
Fig. 4. SEM image of raw 600–630 mm pine stem wood particles (a), sampled at 354 mm (b) and 1062 mm (c) and of raw 600–630 mm wheat straw particles (d), sampled at 354 mm (e) and 1062 mm fall height (f) at 1100 °C DTF temperature.
P. Holmgren et al. / Fuel 206 (2017) 342–351
347
Fig. 5. Optical diameter of the raw fuels factions (a) and Cox circularity (b), determined with the two optical methods SIA and DIA. Error bars indicate 95% confidence interval.
Fig. 6. Riley circularity for pine stem wood plotted in (a) and Cox circularity in (b). For wheat straw Riley circularity is plotted in (c) and Cox circularity in (d).
(Fig. 6a and c), thus failing to describe the marked geometrical differences found. Using Cox circularity instead, the differences both between pine stem wood and wheat straw, and between the different stem wood size fractions was well described (Fig. 6b and d). Riley circularity was thereby concluded less suitable for describing the shape change of the fuels used. Throughout the rest of this paper an assumption of Cox circularity was used. 3.3. Particle size changes By use of the shape descriptors from the previous section, Eq. (3) can be solved giving a volume equivalent diameter for the particles. In the case of stem wood (Fig. 7a) higher temperature leads to faster shrinkage, while thermal inertia of the larger initial size particles slowed down the rate of shrinkage during the first 0.3 s. Initial size did not seem to play a large role for final size in relative terms. Size appeared to decrease between all measurement points. Taking only size into account, this would indicate that, at least for the larger size fractions, the residence time was probably not long
enough for full devolatilization of stem wood particles. Wheat straw particles behaved differently compared to stem wood (Fig. 7), as it reacted much faster and there was no noticeable size change after the first measuring point for any size fraction, shrinking to 59 ± 6% of initial size. However, the smallest sieve size (125– 150 mm) deviated strongly from the other sizes, shrinking to 82 ± 8% of initial size (95% confidence interval). Given the difference in chemical composition of the smallest wheat straw fraction as compared to the larger ones, the higher amount of ash forming elements could contribute to volume stability during devolatilization, thus decreasing the observed shrinkage. Differences in macromolecular composition, possibly caused by fractionation during milling and sieving, may also affect char formation. Describing how the diameter of a particle changes as a function of particle conversion is often described using Eq. (9), which is a slight variation of an equation proposed by Smith [30] for coal fuel particles, but where the particle equivalent diameter is used instead of the measured circle equivalent diameter. X is the degree of conversion, D0eq is the equivalent diameter of the raw fuel, while
348
P. Holmgren et al. / Fuel 206 (2017) 342–351
Fig. 7. Relative diameter change of particles plotted against residence time, pine stem wood in (a) and wheat straw in (b).
the diameter evolution coefficient b can vary between 0 and ⅓. A value of 0 describes a particle where mass loss occurs strictly through density loss while the diameter remains constant, and a value of ⅓ describes a situation where the density remains constant and mass is lost solely through shrinkage. To quantify the degree of conversion density is needed.
Deq ¼ D0eq ð1 XÞb
b¼0
Density loss model
b ¼ 1=3 Shrinking particle model
ð9Þ
conversion proceeds. The term q0geo is the geometric density of the raw fuel while the density evolution coefficients c and d fulfills the same role as b in Eq. (9), though not limited to a maximum value of ⅓. One limitation with this equation is that density at full mass conversion (X = 1) would be equal to raw fuel density. However this model is developed for use on a dry fuel basis, and at X = 1 the particle will have a volume of zero (see Eq. (9)). The char residue and ash composition of observable particles will prevent X from reaching 1.
qgeo ¼ q0geo ðð1 XÞc þ X d Þ 3.4. Particle density changes The weight of the raw fuel particles and char particles was carefully measured and then analyzed by SIA to acquire their volume equivalent diameter. This was translated to total sample volume, and thus average particle density was determined (Eq. (7)). Because of experimental difficulties in accurately weighing and collecting useful images for the medium and smallest fuel fractions, only the 600–630 mm particles were used for density change calculations. Pine stem wood particles experienced an initial decrease in optically determined geometric density as they were devolatilized, followed by an increase in density (Fig. 8a), likely due to the particle undergoing plasticization with subsequent shrinking of pores and internal cavities, compare with Fig. 4(c). Higher temperatures led to a slightly higher initial density drop. This could be explained by higher heating rates leading to a more violent release of volatiles. Density change for wheat straw was initially slow followed by a steep decrease somewhere between 0.3 s and 0.6 s residence time, after which density was stabilized (Fig. 8b), indicating that devolatilization was completed. In this paper, modeling density change during devolatilization is done using a similar approach as for diameter change, though the equation contains one additional term containing X (conversion), see Eq. (10). This allows the density to move from density loss to shrinking particle behavior (or the other way around) as
ð10Þ
Eq. (10) can be extended to cover full conversion to pure ash by the inclusion of a constant before the second X term in the equation, describing the ratio between the density of a pure ash particle and that of a raw fuel particle. 3.5. Fuel conversion model For the large sized fuel fractions geometric density and volume equivalent size were directly measured. By combining relative volume change and relative density change, conversion degree could be assessed and so b, c and d in Eqs. (9) and (10) could be determined for the different fuels. See Table 3 for details. By multiplying the relative volume change with the relative density change of particles, mass loss on a dry basis was assessed by use of an optical conversion model (OCM) according to Eq. (11). Table 3 Empirical constants used to model conversion. Detailed fitting results can be found in Supplementary Material (S3 And S4). Biomass type
Temperature (°C)
b
c
d
Pine stem wood Pine stem wood Wheat straw Wheat straw
900 1100 900 1100
0.33 0.31 0.19 0.19
0.66 0.47 1.48 1.11
1.60 4.15 0.90 3.69
Fig. 8. The relative density change of pine stem wood (a) and wheat straw (b) particles as a function of residence time. Error bars represent 95% confidence interval.
P. Holmgren et al. / Fuel 206 (2017) 342–351
X ¼1
Deq D0eq
0
3 bc 1 Deq b @ Deq þ 1 D0eq D0eq
!d 1 A
ð11Þ
A particle that reaches a size and density change that corresponds to 100% mass conversion was defined as fully converted. The relative densities of the medium and small size fractions could then be estimated based on their relative size, and thus mass conversion could be predicted. The flow scheme is illustrated in Fig. 9. This mass loss was compared with fuel conversion estimated using the reference method, see Fig. 10. Error bars correspond to one standard deviation. 3.5.1. Conversion of pine stem wood particles At 900 °C the OCM gave a slightly but significantly sharper conversion curve for smaller initial particle size, which was expected due to lower thermal inertia and thus higher internal heating rates
Fig. 9. Modified model design flow scheme due to limited density data.
349
for smaller particles. The reference method, on the other hand, indicated no dependence on initial size on conversion as all samples seem to follow a similar curve. One sample analyzed using the reference method had a negative mass loss in Fig. 10(a), likely due to a weighing error during analysis. This illustrates the sensitivity to experimental errors not covered within the analytical uncertainties, and may explain why no size dependence on initial conversion was found. The two methods can also be compared at 1100 °C in Fig. 10(b) where the size dependence is more obvious. Details are available in Supplementary Material (S5). After residence times of approximately 0.25 s the largest stem wood fuel fraction (600–630 mm sieve size) was converted by 30% running the reactor at 1100 °C, while the medium size was converted to around 70%. One standard deviation overlap between the two methods in all cases indicates comparable accuracies.
3.5.2. Conversion of wheat straw particles For wheat straw the differences were consistent at both temperatures, though larger at 900 °C. 600–630 mm sized wheat straw particles gave higher conversion values using OCM compared to the reference values, and lower values for the 125–150 mm fraction (Fig. 10c and Fig. 10d). 400–425 mm particles showed no difference between the methods used. The low conversion estimates of the smallest wheat straw faction at 900 °C was expected due to the way the shrinkage behavior of the smallest wheat straw particles deviated from the other size fractions (see Fig. 7b), however the smallest faction at 1100 °C gave conversion estimates that were in close agreement with the reference data. Details are available in Supplementary Material (S6). According to the optical conversion estimations the initial size of the wheat straw particles did not seem to have an impact on conversion rates within the size range tested. This could be explained by the particle shapes where thickness is the limiting factor for heat transfer, and in this case it is far smaller than the measurements characterizing the particle size. Small differences could be seen both for initial size and the temperature used using the reference method, even though data are a bit scattered.
Fig. 10. Particle conversion plotted against residence time for pine stem wood at 900 °C (a), pine stem wood at 1100 °C (b), wheat straw at 900 °C (c) and wheat straw at 1100 °C.
350
P. Holmgren et al. / Fuel 206 (2017) 342–351
3.5.3. Comparing wood and straw conversion models Considering the much higher ash content of wheat straw, it was unproblematic to gather enough char and thus ash for accurate ICP-AES analysis, whereas some of the stem wood experiments did not yield enough char for successful analysis to be performed. Another aspect on the reference method, using tracking of nonvolatilized ash forming elements, is that it relies on the element not being lost from the fuel particle during conversion. If fragments are formed and the element is not evenly dispersed throughout the particle, non-representative parts may be lost, causing errors in the estimations. This could be the case for the wheat straw particles since they are rather heterogeneous in composition, coming from different parts of the plant, potentially explaining the somewhat scattered trends seen for conversion estimated by the reference method. However, the none-destructive optical method worked comparable or better than the ICP-AES reference method for stem wood, probably due to stem wood particles being well described by the shape factors used. For directly analyzing conversion degree of solid fuel particles this gives an alternative to time-consuming and costly lab analysis of low-ash fuels. A reactor with optical ports would also enable in-situ measurements of conversion at different stages of a reactor. Considering how size and density change in different ways for the different biofuels tested, it is also clear that conversion simulation models for these kinds of fuels need to use different assumptions on how mass loss occurs depending on biomass. For stem wood, a shrinking particle model might work well for describing the size change, though the changes in density seem to be related to the formation of large pores and cavities during the initial rapid release of volatiles. For wheat straw, on the other hand, conversion seemed to occur in two steps; first through shrinking, and then as density loss. The wheat straw particles never approached a shape that is easily quantified by traditional shape descriptors such as aspect ratio and diameter, and shape aspects therefore requires further studies.
Combining volume and density information, the conversion model was formulated. The modeled conversions were compared with those estimated by a method based on chemical analysis of samples, and a mass balance based on assumptions of some of the ash forming elements not being volatilized during conversion. In the case of pine stem wood, the conversion estimates were in close agreement with values provided by the reference method. Besides accurate description of the transformation, it gave information indicating that thermal inertia limitations limited conversion of the largest size fraction used. Describing volume and density change of wheat straw particles was less accurate, mainly due to difficulties in modeling size and shape change of the very irregularly shaped particles. A strong size dependence in shrinking behavior was found, possibly due to differences in chemical composition originating from the milling and sieving. The optical conversion model developed can be used for multiphase flow modeling purposes or for in-situ monitoring of particle conversion in thermochemical reactors if optical access is provided. For predictions on mass basis, accurate estimations of density is needed, e.g. as described in this paper. Acknowledgement The authors would like to thank the Swedish Energy Agency through the Swedish Centre for Biomass Gasification (SFC), and the Kempe foundation for financial support. The authors would also like to thank Bio4Energy (B4E), a strategic research environment appointed by the Swedish government, for supporting this work. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.fuel.2017.06.009. References
4. Conclusions Size and shape changes of biomass particles during high heating rate devolatilization were studied in the pyrolysis regime using a drop tube furnace, extractive sampling and image analysis. Besides detailed transformation descriptions of the two fuels studied (pine wood and wheat straw powders), a conversion model was developed based on optical 2D data. Accurate models require useful shape descriptors to give simplified but yet correct representations of the shapes analyzed. The two fuels studied had very different initial particle properties, and the differences were even more pronounced during conversion where the pine wood particles formed plasticized hollow spheres, whereas the flake-like wheat straw particles curled and did not shrink as much. The model developed in this paper uses a particle diameter based on 2D particle area corrected by use of circularity based on measured perimeter of a particle. For this purpose, common shape descriptors were evaluated and Cox circularity (the ratio between the perimeter of a circle and the real shape perimeter, assuming equal shape area) was found best suited for describing the shape change of irregularly shaped biomass and char particles. Besides the volumetric change, a mass based conversion requires information on particle density. The density was estimated by weighting and approximating the volume of representative particles. From the density trends, it was found that the density of pine stem wood particles remain rather constant during conversion, whereas wheat straw char had significantly lower density comparted to the original fuel.
[1] Ahmad AA, Zawawi NA, Kasim FH, Inayat A, Khasri A. Assessing the gasification performance of biomass: A review on biomass gasification process conditions, optimization and economic evaluation. Renew Sust Energ Rev 2016;53:1333–47. [2] Dai JJ, Cui HP, Grace JR. Biomass feeding for thermochemical reactors. Prog Energy Combust 2012;38:716–36. [3] Li SG, Xu SP, Liu SQ, Yang C, Lu QH. Fast pyrolysis of biomass in free-fall reactor for hydrogen-rich gas. Fuel Process Technol 2004;85:1201–11. [4] Dupont C, Boissonnet G, Seiler J-M, Gauthier P, Schweich D. Study about the kinetic processes of biomass steam gasification. Fuel 2007;86:32–40. [5] Lu H, Ip E, Scott J, Foster P, Vickers M, Baxter LL. Effects of particle shape and size on devolatilization of biomass particle. Fuel 2010;89:1156–68. [6] Davidsson KO, Pettersson JBC. Birch wood particle shrinkage during rapid pyrolysis. Fuel 2002;81:263–70. [7] Tolvanen H, Kokko L, Raiko R. Fast pyrolysis of coal, peat, and torrefied wood: Mass loss study with a drop-tube reactor, particle geometry analysis, and kinetics modeling. Fuel 2013;111:148–56. [8] Saastamoinen J, Aho M, Moilanen A, Sorensen LH, Clausen S, Berg M. Burnout of pulverized biomass particles in large scale boiler - Single particle model approach. Biomass Bioenergy 2010;34:728–36. [9] Senneca O, Urciuolo M, Chirone R, Cumbo D. An experimental study of fragmentation of coals during fast pyrolysis at high temperature and pressure. Fuel 2011;90:2931–8. [10] Yu JL, Strezov V, Lucas J, Wall T. Swelling behaviour of individual coal particles in the single particle reactor. Fuel 2003;82:1977–87. [11] Zhang LA, Binner E, Chen LG, Qiao Y, Li CZ, Bhattacharya S, et al. Experimental investigation of the combustion of Bituminous coal in air and O-2/CO2 mixtures: 1 particle imaging of the combustion of coal and char. Energy Fuels 2010;24:4803–11. [12] Hodge EM, Roberts DG, Harris DJ, Stubington JF. The Significance of Char Morphology to the Analysis of High-Temperature Char-CO2 Reaction Rates. Energy Fuels 2010;24:100–7. [13] Costa FF, Costa M. Particle fragmentation of raw and torrefied biomass during combustion in a drop tube furnace. Fuel 2015;159:530–7. [14] Zhang Y, Kajitani S, Ashizawa M, Miura K. Peculiarities of rapid pyrolysis of biomass covering medium- and high-temperature ranges. Energy Fuels 2006;20:2705–12.
P. Holmgren et al. / Fuel 206 (2017) 342–351 [15] Dupont C, Chen L, Cances J, Commandre JM, Cuoci A, Pierucci S, et al. Biomass pyrolysis: Kinetic modelling and experimental validation under high temperature and flash heating rate conditions. J Anal Appl Pyrol 2009;85:260–7. [16] Commandre JM, Lahmidi H, Salvador S, Dupassieux N. Pyrolysis of wood at high temperature: The influence of experimental parameters on gaseous products. Fuel Process Technol 2011;92:837–44. [17] Jarvis MW, Haas TJ, Donohoe BS, Daily JW, Gaston KR, Frederick WJ, et al. Elucidation of biomass pyrolysis products using a laminar entrained flow reactor and char particle imaging. Energy Fuels 2011;25:324–36. [18] Dupont C, Commandre JM, Gauthier P, Boissonnet G, Salvador S, Schweich D, et al. Biomass pyrolysis experiments in an analytical entrained flow reactor between 1073 K and 1273 K. Fuel 2008;87(1073):1155–64. [19] Umeki K, Kirtania K, Chen LG, Bhattacharya S. Fuel particle conversion of pulverized biomass during pyrolysis in an entrained flow reactor. Ind Eng Chem Res 2012;51:13973–9. [20] Tamaki Y, Mazza G. Measurement of structural carbohydrates, lignins, and micro-components of straw and shives: Effects of extractives, particle size and crop species. Ind Crop Prod 2010;31:534–41. [21] Molinder R, Wiinikka H. Feeding small biomass particles at low rates. Powder Technol 2015;269:240–6.
351
[22] ImageJ (Software), in, http://imagej.nih.gov/ij/. [23] Bonadonna GGJEC, Sparks RSJ. Thickness variations and volume estimates of tephra fall deposits. J Volcanol Geoth Res 1998;81. [24] Bagheri GH, Bonadonna C, Manzella I, Vonlanthen P. On the characterization of size and shape of irregular particles. Powder Technol 2015;270:141–53. [25] Cox EP. A method of assigning numerical and percentage values to the degree of roundness of sand grains. J Paleontol 1927;179–183. [26] Riley NA. Projection sphericity. J Sediment Res 1941;94–95. [27] Borrego AG, Alvarez D. Comparison of chars obtained under oxy-fuel and conventional pulverized coal combustion atmospheres. Energy Fuels 2007;21:3171–9. [28] Shah KV, Cieplik MK, Betrand CI, van de Kamp WL, Vuthaluru HB. A kineticempirical model for particle size distribution evolution during pulverised fuel combustion. Fuel 2010;89:2438–47. [29] Li S. Char-Slag Transition During Pulverized Coal Gasification Ph.D Thesis. Department of Chemical Engineering, The University of Utah; 2010. p. 73. [30] Smith IW, The Combustion Rates of Coal Chars: A Review, Nineteenth Symposium (International) on Combustion/The Combustion Institute, (1982).