2.1. Characterization of the Coffee Ground Composition
For modeling the thermal degradation of biomass particle, it is important to determine the biomass composition because biomass consists of various species which show different pyrolyzing pathways. The major components which constitute biomass are generally cellulose, hemicellulose and lignin. Extractive species, protein and ash can be other minor components. But detailed information of biochemical composition which can provide the relative amount of these components is not commonly available because there still remain difficulties and uncertainties in current measurement techniques. To overcome this problem, an alternative approach has been suggested by Ranzi et al. [
23] to derive the biomass composition in an indirect way based on the C/H/O atomic balance of biomass. In the previous study on the sawdust pyrolysis in a tilted-slide reactor [
31], five reference components have been adopted to represent cellulose, hemicellulose and three different types of lignins.
Figure 1 shows these reference species in the H/C diagram. The reference components CELL represent the cellulose monomer (C
6H
10O
5). The hemicellulose is represented as a reference component HECELL which is approximated to be a xylose monomer (C
5H
8O
4). Because lignin has a complex chemical structure, three reference components of LIG-C (C
15H
14O
4), LIG-H (C
22H
28O
9) and LIG-O (C
20H
22O
10) were adopted which are richer in C, H and O, respectively. The three reference mixtures, RM-1, RM-2 and RM-3 are specified to reduce the degree of freedom in the biomass characterization. These reference mixtures are defined by the splitting parameters,
α,
β and
γ. The mixture RM-1 is composed of 60% CELL and 40% HECELL in molar ratio where
α is defined to be 0.6. RM-2 and RM-3 are composed of 80% LIG-H and 20% LIG-C (
β = 0.8) and 80% LIG-O and 20% LIG-C (
γ = 0.8), respectively. These reference mixtures RM-1, RM-2 and RM-3 form the vertices of the dark-shadowed triangle of
Figure 1. When the elemental composition of biomass is placed inside this triangle, its biochemical composition can be derived from a linear combination of RM-1, RM-2 and RM-3. The sawdust biomass in the previous study falls within the triangle as represented to be ‘SD’ in
Figure 1 and the five reference components are sufficient for biomass characterization.
The coffee grounds have been collected from coffee shops in Seoul, South Korea. The moisture content of collected coffee ground was higher than 50 wt% and it should be dried prior to fast pyrolysis. The coffee ground was dried in a rotary dryer to lower its moisture content down to less than 10 wt%. We note that the drying process only evaporates the water in coffee ground without changing its chemical structure and the drying pretreatment could only affect the water content in bio-crude oil.
Table 1 shows the result of proximate analysis for coffee ground. The proximate analysis was performed using a thermogravimetric analyzer (LECO TGA 701). Because the moisture and ash contents are not negligible in coffee ground, their effects were included in the simulation which will be discussed later in detail.
Table 2 shows the elemental composition of coffee ground. The elemental composition was measured by an elemental analyzer (Flash EA 1112 series). For the biomass characterization, the C/H/O ratios were normalized to make their sum to be 100% in dry and ash-free basis. The elemental composition of coffee ground indicated by ‘CG’ in
Figure 1 falls outside the dark-shadowed triangle formed by RM-1, RM-2 and RM-3. It can be known that coffee ground has larger C and H contents in comparison with sawdust and it cannot be represented by the five reference species (CELL, HECELL, LIG-C, LIG-H and LIG-O). It is known that several kinds of softwoods, grass plants and seeds are relatively rich in extractive species and the area of biomass characterization needs to be extended by introducing new reference components [
26]. Because there is a great variety of the extractive species, a couple of lumped species were introduced for simplification depending on their water solubility, that is, hydrophobic and hydrophilic extractives. For hydrophobic extractives, triglycerides which have large amount of linoleic acid are most common. Based on this, a reference component named TGL (C
57H
100O
7) with large C and H content was selected [
26] as marked in
Figure 1. For hydrophilic species, phenolic compounds are typical which can be represented by tannin species. The component named TANN (C
15H
12O
7) with a small H amount was selected [
26] as indicated in
Figure 1.
The biomass characterization can be extended significantly by introducing new species of TGL and TANN. The inclusion of TGL in RM-2 can cover the biomass richer in hydrogen, while the biomass with lower hydrogen content can be characterized by the inclusion of TANN in RM-3. The new splitting parameter
δ defines the molar ratio of lignins (LIG-H and LIG-C) in RM-2, which also means that the molar ratio of TGL in RM-2 is 1-
δ. Another splitting parameter
ε defines the molar ratio of lignins (LIG-O and LIG-C) in RM-3 which means that the molar ratio of TANN in RM-3 is 1-
ε. Because coffee ground is richer in carbon and hydrogen, it is enough to migrate the vertex RM-2 in the direction of TGL, while the location of RM-3 need not to be modified. The splitting parameters
α/
β/
γ/
δ/
ε = 0.6/0.8/0.8/0.7/1 can successfully extend the characterization region to light-shadowed area to cover coffee ground as shown in
Figure 1. The biochemical composition is then derived as shown in
Table 3 by adding moisture and ash which were determined from the proximate analysis.
2.2. Simulation Method
The fast pyrolysis of coffee ground had been performed in a tilted-slide reactor [
29]. The schematic diagram of the reactor system is shown in
Figure 2. The biomass is supplied via a screw conveyor into the top of the tilted reactor. The heated sand is also supplied to the top of the reactor as a heat transfer medium for fast pyrolysis. The biomass is pyrolyzed while it is transported down together with hot sand and their movements are guided baffles installed on the reactor floor. The pyrolysis products are gaseous species and residual char. The gaseous products are composed of condensable volatiles and noncondensable gases, which pass though the condensers where the volatiles are condensed into bio-crude oil. The solid residue is biochar and it is conveyed with sand from the reactor bottom to the circulating fluidized bed combustor where the biochar is burned to heat up the sand. In the tilted-slide reactor, the oxygen-free condition for the fast pyrolysis can be maintained due to the sand filled in the sand hopper above the reactor and that filled in the screw conveyer below the reactor. In this study, as the simulation region the tilted reactor is focused on and the condensation of volatile was not accounted for. Therefore, it is assumed that the yield of bio-crude oil is equivalent to that of volatile. The simulation is carried out using a commercial computational fluid dynamics software (STAR-CCM+ v9.06). The geometry for the simulation is shown in
Figure 3 together with the trimmer mesh with rectangular cuboid shape. The meshes were composed of 40,113 cells. During the experiment, the reactor is operated at a slightly negative pressure condition by the induced draft fan to prevent a leak of pyrolytic gas from the reactor. The effect of induced draft fan is implemented by introducing a fan interface as indicated in
Figure 3. The fan performance curve provided by the fan manufacturer is applied to the simulation and the fan rotation rate is specified at 286 rpm which corresponds to 5 Hz of the inverter to control the fan rotation rate in the experiment.
The governing equations of continuity, momentum, energy and species equations were discretized by finite volume method and the segregated approach was used to solve the flow equations. The details of governing equations and solving procedure are described in previous study [
31] and STAR-CCM+ user guide [
34]. To simulate the behaviors of sand and biomass particles, a Lagrangian multiphase model is adopted. The densities of sand and coffee ground were 2650 and 1145 kg/m
3, respectively. The specific heat of sand is evaluated by a polynomial as a function of temperature to be
. The specific heats of coffee ground and biochar are treated separately, which are calculated to be
and
, respectively. The residence time of sand particle was measured to be 4.5s at the atmospheric condition. In the simulation, the normal and tangential restitution coefficients of particles are selected to be 0.8 by trial and error to have a similar residence time.
The fluid-particle heat transfer is calculated using the Nusselt number with the expression from Gunn as follows [
35]:
where
is the volume fraction of gas-phase. It is known that this expression is more suitable for a wide range of particle volume fractions. The particle-particle heat transfer is not accounted for in this study. The fluid-particle heat transfer coefficient
was evaluated as follows:
where
and
are the thermal conductivity of gas-phase and the particle diameter, respectively.
The kinetics of coffee ground pyrolysis are described by a lumped kinetic model [
27] simplified from detailed kinetics of cellulose, hemicellulose and lignin decomposition [
36,
37,
38,
39]. The reactions and their kinetic parameters have been progressively extended and updated based on the experimental data [
27] and the model predictions adopting the pyrolysis kinetics have been verified in various reactor configurations including a drop tube reactor [
23], a fixed-bed reactor [
24], a fluidized bed reactor [
15,
24,
32], a conical spouted bed reactor [
32] operating in various fast pyrolysis conditions such as heating rates, reaction temperatures and gas and solid residence times. It has been shown that the maximum product yields as wells as the yield trends with temperature could be reasonably predicted.
It was assumed that each reference component decomposes through first-order reactions independently. The kinetic model includes 22 solid species and 28 gas species which constitute 25 reactions. The solid and gaseous species are listed in
Table 4 and
Table 5, respectively. The kinetic parameters are summarized in
Table 6.
It is known that the ash content in the biomass can decrease the bio-crude oil yield by catalytic effect [
27]. The major ash effect is the reduction of levoglucosan and xylan in favor of decomposition and dehydration products along with a char increase. The experimental data obtained in the VTT’s pyrolyzer [
40] showed that higher ash content leads to lower oil yield with higher water content. The decrease of oil yield as a function of ash content becomes almost negligible as the ash content are larger than 4 wt% [
40]. To account for the ash effect, the simplified approach [
27] was adopted and the kinetic parameters of a few selected reactions have been modified to maintain a flexible and simple mechanism. The reduction of levoglucosan and xylan along with char increase were taken into account by introducing a global ash factor as follows [
27]:
It was assumed that catalytic effect is already completed for ash content of 5 wt% [
27] and the ash factor (AF) was defined as Equation (3) where the value of AF ranges from zero to asymptotically 1. In this study, the ash content is 2.77% in coffee ground which leads to the ash factor of 0.882. The activation energies of active cellulose decomposition (
) and charification reaction (
) are modified using AF as follows:
The activation energies of the decomposition of hemicellulose intermediate (HCE1) are modified as follows:
The feeding rate of coffee ground is 17.7 kg/h and the sand flow rate is 662.6 kg/h. For computational efficiency, instead of tracking individual particles a parcel stream is adopted as a group of particles of same velocity. Per injection points of sand and coffee ground, the numbers of parcel streams are set at 100. The temperature distribution inside the particle is assumed to be uniform. The particle size distributions of sand and coffee ground are shown in
Figure 4 as cumulative distribution functions (CDF).
The simulation of the fast pyrolysis in a tilted-slide reactor is performed at various temperatures in a steady-state. The temperatures of injecting sands are 500, 550, 600, 650 °C whereas the biomass temperature entering the reactor is 25 °C. The “reactor temperature” is determined at a specified location in the reactor as indicated in
Figure 3, which corresponds to the thermocouple position in the experiment. The reactor temperatures will be used in the comparison with the experimental results and further analysis.