Simulation of a Fluidized Bed Dryer for the Drying of Sago Waste

: The large amount of sago waste produced by sago processing industries can cause serious environmental problems. When dried, these residues usually have a high starch content (around 58%) and have many potential applications. In this study, the drying of sago waste using a ﬂuidized bed dryer (FBD), which offers more advantages than other drying methods, is analyzed via computational ﬂuid dynamics (CFD) modeling. A two-dimensional (2D) FBD model is also developed and a mesh independency test is conducted immediately afterwards. A ﬁne mesh is selected for the CFD model and a simulation is conducted using ANSYS Fluent 17.1 software (Ansys Inc., version 17.1, Canonsburg, PA, USA). The governing and discretized algebraic equations are solved by applying the phase-coupled semi-implicit method for pressure-linked equations. Both the Eulerian–Eulerian multiphase model approach and the turbulence model are applied in the simulation due to the turbulent ﬂow in the dryer. A velocity of 1.30 m/s and temperature of 50 ◦ C are selected as boundary conditions based on the optimum parameter values from previous experiments. The ﬁnal moisture content that we aim to achieve is 10% or a moisture ratio of 0.25 in sago waste for the purpose of animal feed, so as to prevent bacterial growth and for packaging purposes based on common industrial practice. Both the drying rate and ﬂuidization proﬁle are examined at air velocities of 0.6, 1.0, 1.3, 1.8, and 2.2 m/s. Based on the results, the velocity range of 1.0 m/s to 2.2 m/s is deemed suitable for the ﬂuidization and drying of sago waste with a particle size of 2000 µ m for a drying simulation of 1 h. The drying rate is further examined at air temperatures of 50 ◦ C, 60 ◦ C, 70 ◦ C, and 80 ◦ C, whereas the ﬂuidization proﬁle is examined at particle sizes of 200, 500, 1000, and 2000 µ m. The results reveal excellent ﬂuidization at a particle size range of 500 µ m to 2000 µ m and a velocity of 1.3 m/s. L.P.C.; Software, M.I.R. and L.P.C.; Validation, M.I.R.; Formal analysis, L.P.C.; Investigation, L.P.C.; Resources, M.I.R. and M.S.T.; Data Curation, L.P.C.; Writing-original draft, L.P.C.; Writing-review & editing, M.I.R., A.M.A.N. and M.S.T.; Visualization, M.I.R. and A.M.A.N.; Supervision, M.I.R.; Project Administration, M.I.R. and M.S.T.; Funding Acquisition, M.I.R. and M.S.T.


Introduction
Sago palm (Metroxylon spp.) is a tropical plant that can adapt well in wet growing conditions [1]. Malaysia, especially Sarawak, is the largest sago exporter in the world and currently has nine sago processing industries that operate actively and export around 25,000 to 40,000 tons of sago products to Peninsular Malaysia, Japan, Taiwan, Singapore, and other countries [2]. However, the extraction of starch and the large-scale production of sago also generate sago waste or pith residues. Sarawak, especially its towns Sibu and Mukah, produces a large amount (at least 50 to 110 tons) of sago waste every day from their sago processing industries. These residues comprise 58% starch, 23% cellulose,

Methodology
Numerical simulation required the type of model, properties of parameters on each boundary, and the operating conditions. Several assumptions were made and the governing equations in the modeling process were formulated. Meshing analysis was then performed and the results were validated by comparing them with the experimental data. The simulation was conducted using ANSYS Fluent 17.1 installed on a computer with an Intel Core i5 processor, 8.00 GB RAM, and a Windows 8 64-bit operating system.

Development of Geometry and Mesh
The geometry dimensions for this work (0.22 m in height and 0.18 m in diameter) were borrowed from Ho [17] and constructed using ANSYS DesignModeler 17.1 (Ansys Inc., version 17.1, Canonsburg, PA, USA). The sago waste particles that were used as bed materials in the dryer had a height of 0.02283 m, which was calculated based on the density (1033 kg/m 3 ) and mass (600 g) of sago waste in the experimental study.
A mesh independency test was conducted after developing the geometry. A fine mesh with 5141 elements and a minimum size of 0.039672 mm was used. Figure 1 shows the result of the meshing process.

Assumptions
Some assumptions were made for the simulation: 1. No chemical reaction takes place in the drying process. 2. The gas and solid are well-mixed. 3. No slip conditions are set. 4. The sago waste has an initial moisture content of 40%, as this is the suitable range of moisture content inlet for drying in a fluidized bed dryer. 5. The sago waste does not agglomerate at a moisture content of 40%. 6. The desired final moisture content of sago waste at the fluidized bed dryer outlet is selected to be 10% to preserve the nutrients of the sago waste [18]. 7. Particle size is not reduced with drying time and the different particle sizes are classified based on Geldart theory. 8. The thermal properties of sago are assumed to be the same as those of cassava.

Assumptions
Some assumptions were made for the simulation:

1.
No chemical reaction takes place in the drying process. 2.
The gas and solid are well-mixed.
The sago waste has an initial moisture content of 40%, as this is the suitable range of moisture content inlet for drying in a fluidized bed dryer. 5.
The sago waste does not agglomerate at a moisture content of 40%. 6.
The desired final moisture content of sago waste at the fluidized bed dryer outlet is selected to be 10% to preserve the nutrients of the sago waste [18]. 7.
Particle size is not reduced with drying time and the different particle sizes are classified based on Geldart theory. 8.
The thermal properties of sago are assumed to be the same as those of cassava.

Modeling
The Eulerian-Eulerian multiphase model is used in this study because this tool requires less computational effort and is widely used in FBDs for CFD modeling. This model is composed of the air as the gas phase and sago waste as the solid phase. The heat transfer model is activated to analyze the heat transfer at the gas and solid regions in the model. A turbulence model is also employed due to the turbulent flow in the dryer. The standard k-ε model is also activated due to its simplicity and minimal computational requirements. In order to obtain the moisture ratio, model two-term exponential is inserted as a user-defined function.

Governing Equation
The conservations of mass and momentum are the main equations considered in the multiphase model. The conservation of energy and the equations in the turbulence model must also be considered for the heat transfer and the turbulence flow.
Conservation of mass [19]: where: → v q = velocity of q phase; . m pq = mass transfer from p th phase to q th phase; . m qp = mass transfer from q th phase to p th phase.
Conservation of momentum [19]: where: = τ q = q th phase shear stress tensor; u q = shear viscosity of q phase; λ q = bulk viscosity of q phase; → F q = external body forces; → F li f t,q = lift forces; → F wl,q = wall lubrication forces; → F td,q = virtual mass and turbulent dispersion forces; → R pq = interaction force between phases; p = pressure shared by all phases.

Conservation of energy [19]
∂ ∂t α q p q h q + ∇. α q p q where: h q = specific enthalpy of q th phase; → q q = heat flux; S q = source of enthalpy; Q pq = intensity of heat exchange between p th and q th phases; h pq = interphase enthalpy.
Transport equation [19]: where: G k = generation of turbulence kinetic energy due to mean velocity gradients; G b = generation of turbulence kinetic energy due to buoyancy; Y M = contribution of the fluctuating dilatation in compressible turbulence to overall rate; C 1 , C 2 , and C 3 = constants; S , S k = user source term.

Operating and Boundary Conditions
The gravitational acceleration was set to 9.81 m/s for the operating condition. Moreover, the inlet velocity of 1.30 m/s, temperature of 50 • C, and particle size of 2000 µm were used as boundary conditions. The inlet velocity of solid was set to 0 m/s to fix the bed in its original position. A no-slip condition was also set for the wall. The solid used in the simulation had a volume fraction of 0.45 [13]. The pressure outlet was also selected to reduce the reversed flow during the simulation.

Solution Procedure
The governing equations were solved by choosing the appropriate boundary conditions. The first-order implicit was selected; this is considered sufficient for the transient formulation in most problems. The first-order upwind was used in the spatial discretization for the convective terms except for the momentum, whilst the second-order upwind was used to improve the convergence. The phase-coupled semi-implicit method for pressure-linked equations (SIMPLE) was used for pressure velocity coupling in the multiphase model. A time step value of 0.075 s was also used to guarantee the stability of each simulation. Each simulation was run for 1 h with 48,000 time steps.

Validation of the Model
The model was validated by comparing the simulation results with the experimental findings of Ho (2016) [17] at an inlet air velocity of 1.30 m/s, inlet air temperature of 50 • C, and particle size of 2000 µm. The simulation results of this two-term exponential model were also compared with the simulation results of Zahir (2016) [20], as shown in Figure 2. Both the simulation and experimental data curve showed a diverse trend with the spherical particle data due to the different utilizations of the tool and model by the researcher. The simulation curve exhibited the same trend as the experimental data, thereby generating a minimal number of errors in the comparison. This simulation work also showed an improved prediction of the drying profile for sago waste compared to the simulation work performed by Zahir using a model of spherical particles.   Figure 3 shows that the moisture ratio curve continuously decreases with time. Meanwhile, the drying time it takes to achieve a final moisture ratio of 0.25 decreases along with increasing air velocity. The shortest required drying time (56 min) is recorded at a peak velocity of 2.2 m/s but exceeds 1 h at a velocity of 0.6 m/s. This finding can be ascribed to the fact that a higher heat transfer rate is recorded at higher velocities. Specifically, the drying time reaches 59, 58, and 57 min at velocities of 1.0, 1.3, and 1.8 m/s, respectively. These results clearly indicate that velocity does not have any significant effect on the moisture ratio curve because the differences amongst the drying times recorded at velocities of 1.0, 1.3, 1.8, and 2.2 m/s are only 1 min at most. This finding may be attributed to the fact that the considered velocity range is not large enough to produce an obvious effect.  Figure 3 shows that the moisture ratio curve continuously decreases with time. Meanwhile, the drying time it takes to achieve a final moisture ratio of 0.25 decreases along with increasing air velocity. The shortest required drying time (56 min) is recorded at a peak velocity of 2.2 m/s but exceeds 1 h at a velocity of 0.6 m/s. This finding can be ascribed to the fact that a higher heat transfer rate is recorded at higher velocities. Specifically, the drying time reaches 59, 58, and 57 min at velocities of 1.0, 1.3, and 1.8 m/s, respectively. These results clearly indicate that velocity does not have any significant effect on the moisture ratio curve because the differences amongst the drying times recorded at velocities of 1.0, 1.3, 1.8, and 2.2 m/s are only 1 min at most. This finding may be attributed to the fact that the considered velocity range is not large enough to produce an obvious effect. Figure 4 shows the solid volume fraction contour at different velocities. These contours exhibit the same trends across different bed heights in a simulation time range of 0 s to 10 s.

Effect of Air Velocity on CFD Modeling Performance
At 0 s, the solid bed is fixed at a height of 0.02283 m with a velocity of 0 m/s. This bed starts to fluidize at 1 s and demonstrates a fluid-like behavior when the inlet upward-flowing gas interacts with the particle bed. The drag force, which refers to the frictional force imposed by the inlet air and downward gravitational force, is addressed. At the initial stage, the bed expands rapidly until reaching its maximum height. At a velocity of 0.6 m/s, both fluidization and bed expansion are kept at minimal levels, whereas the recorded bed height is almost the same as its initial level.
Energies 2018, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/energies exceeds 1 h at a velocity of 0.6 m/s. This finding can be ascribed to the fact that a higher heat transfer rate is recorded at higher velocities. Specifically, the drying time reaches 59, 58, and 57 min at velocities of 1.0, 1.3, and 1.8 m/s, respectively. These results clearly indicate that velocity does not have any significant effect on the moisture ratio curve because the differences amongst the drying times recorded at velocities of 1.0, 1.3, 1.8, and 2.2 m/s are only 1 min at most. This finding may be attributed to the fact that the considered velocity range is not large enough to produce an obvious effect.  At 0 s, the solid bed is fixed at a height of 0.02283 m with a velocity of 0 m/s. This bed starts to fluidize at 1 s and demonstrates a fluid-like behavior when the inlet upward-flowing gas interacts with the particle bed. The drag force, which refers to the frictional force imposed by the inlet air and downward gravitational force, is addressed. At the initial stage, the bed expands rapidly until reaching its maximum height. At a velocity of 0.6 m/s, both fluidization and bed expansion are kept at minimal levels, whereas the recorded bed height is almost the same as its initial level. At 3 s, the bed region is flattened as the bed height decreases from the level recorded at 1 s. According to Verma et al. [21], the reduction in bed height before the flattening of the bed region could be ascribed to the formation of huge bubbles at 1 s.
At 10 s, the bed region is flattened at each velocity and remains in its flat state despite the increase in bed height.

Effect of Particle Size on CFD Modeling Performance
The effects of particle sizes 200, 500, 1000, and 2000 µm on the contour solid volume fraction are shown in Figure 6. These contours are divided into two phases, where the red color represents the solid phase with a volume fraction of 0.45, while the blue color represents the air phase with a volume fraction of 0. Sago has a density of 1033 kg/m 3 . Moreover, according to Geldart theory [22],  At 3 s, the bed region is flattened as the bed height decreases from the level recorded at 1 s. According to Verma et al. [21], the reduction in bed height before the flattening of the bed region could be ascribed to the formation of huge bubbles at 1 s. At 10 s, the bed region is flattened at each velocity and remains in its flat state despite the increase in bed height. Figure 5 shows that the bed height reaches 0.025, 0.03, 0.035, 0.0425, and 0.05 m at velocities of 0.6, 1.0, 1.3, 1.8, and 2.2 m/s, respectively. At each velocity, the bed height slightly increases from its initial level of 0.00283 m because those particles with a size of 2000 µm cannot be easily fluidized according to Geldart theory.

Effect of Particle Size on CFD Modeling Performance
The effects of particle sizes 200, 500, 1000, and 2000 µm on the contour solid volume fraction are shown in Figure 6. These contours are divided into two phases, where the red color represents the solid phase with a volume fraction of 0.45, while the blue color represents the air phase with a volume fraction of 0. Sago has a density of 1033 kg/m 3 . Moreover, according to Geldart theory [22], outside of the bed. Thus, the air velocity has a more significant effect on fluidization compared to the drying rate. The results showed that a velocity range of 1.0 m/s to 2.2 m/s rendered the particles with a size of 2000 µm able to fluidize.

Effect of Particle Size on CFD Modeling Performance
The effects of particle sizes 200, 500, 1000, and 2000 µm on the contour solid volume fraction are shown in Figure 6. These contours are divided into two phases, where the red color represents the solid phase with a volume fraction of 0.45, while the blue color represents the air phase with a volume fraction of 0. Sago has a density of 1033 kg/m 3 . Moreover, according to Geldart theory [22], those particles with sizes of 200, 500-1000, and 2000 µm are classified into groups A, B, and D, respectively. Particle group C with a size range of 10-20 µm was not included since it is not suitable to be used in FBD and shows poor fluidization [23].  In Figure 6a, it is shown that rapid fluidization takes place at 0.15 s and that the fluidization process continues until all particles are entrained out of the bed at 0.3 s. Entrainment of the solid particles causes a loss in particles mass. The inlet gas also continues to impose a very high drag force until the equal and opposite drag forces imposed by the particles are fully overcome. In Figure 6a, it is shown that rapid fluidization takes place at 0.15 s and that the fluidization process continues until all particles are entrained out of the bed at 0.3 s. Entrainment of the solid particles causes a loss in particles mass. The inlet gas also continues to impose a very high drag force until the equal and opposite drag forces imposed by the particles are fully overcome.
The largest bed expansion is observed amongst particles with a size of 500 µm, as seen in Figure 6b. In this case, the fluidization begins at 0.15 s and the bed further expands at 1 s due to the behavior of the particles in group B, which are easily fluidized. According to Geldart [22], the particles in this group often undergo bubbling fluidization rather than smooth fluidization. Given that bubbles are formed at the beginning of the fluidization at 0.15 s, the minimum bubbling velocity is the same as the minimum fluidization velocity.
In Figure 6c, similar findings are observed for those particles with a size of 1000 µm, which are also classified under group B. Meanwhile, the bed only shows a minimal expansion for those particles with a size of 500 µm because at a velocity of 1.3 m/s, these particles show a less smooth fluidization compared with the other particles. Those particles with a size of 2000 µm (group D) undergo bubbling fluidization with minimal bed expansion. Moreover, the formed bubbles are unable to reach their maximum size and tend to fall back to the lower bed height after 10 s. These particles also show a poor fluidization compared with the particles in group B. Figure 7 shows the expansion of the bed across different particle sizes from its initial height of 0.02283 m. A smaller particle size shows a greater bed expansion because small particles can be easily fluidized. However, fluidization also depends on air velocity, as can be observed in the fluidization of particles with a size of 200 µm. As can be seen in Figure 7, those particles with sizes of 500, 1000, and 2000 µm demonstrate the greatest bed expansions of 0.125, 0.075, and 0.035 m, respectively. The bed height for 200-µm particles was excluded as the value of the bed during fluidization exceeded the height of the dryer. Therefore, at a velocity of 1.3 m/s, the ideal particles for the fluidization had a size range of 500 µm to 2000 µm. The particles in group B exhibited better fluidization than those in group D. The particles in group A are not recommended for the velocity of 1.3 m/s because such a velocity will entrain all these particles out of the bed.

of 13
The largest bed expansion is observed amongst particles with a size of 500 µm, as seen in Figure 6b. In this case, the fluidization begins at 0.15 s and the bed further expands at 1 s due to the behavior of the particles in group B, which are easily fluidized. According to Geldart [22], the particles in this group often undergo bubbling fluidization rather than smooth fluidization. Given that bubbles are formed at the beginning of the fluidization at 0.15 s, the minimum bubbling velocity is the same as the minimum fluidization velocity.
In Figure 6c, similar findings are observed for those particles with a size of 1000 µm, which are also classified under group B. Meanwhile, the bed only shows a minimal expansion for those particles with a size of 500 µm because at a velocity of 1.3 m/s, these particles show a less smooth fluidization compared with the other particles. Those particles with a size of 2000 µm (group D) undergo bubbling fluidization with minimal bed expansion. Moreover, the formed bubbles are unable to reach their maximum size and tend to fall back to the lower bed height after 10 s. These particles also show a poor fluidization compared with the particles in group B. Figure 7 shows the expansion of the bed across different particle sizes from its initial height of 0.02283 m. A smaller particle size shows a greater bed expansion because small particles can be easily fluidized. However, fluidization also depends on air velocity, as can be observed in the fluidization of particles with a size of 200 µm. As can be seen in Figure 7, those particles with sizes of 500, 1000, and 2000 µm demonstrate the greatest bed expansions of 0.125, 0.075, and 0.035 m, respectively. The bed height for 200-µm particles was excluded as the value of the bed during fluidization exceeded the height of the dryer. Therefore, at a velocity of 1.3 m/s, the ideal particles for the fluidization had a size range of 500 µm to 2000 µm. The particles in group B exhibited better fluidization than those in group D. The particles in group A are not recommended for the velocity of 1.3 m/s because such a velocity will entrain all these particles out of the bed.

Conclusions
A simulation was performed to examine the influence of inlet air velocity, inlet air temperature, and particle size on achieving the final moisture ratio of 0.25. The multiphase Eulerian model was used to describe the gas-solid turbulent flow. A 2D CFD model with a fine mesh was also developed. The small errors in the model validation indicated that this model is acceptable. The optimum parameters (i.e., 1.3 m/s, 50 °C, and 2000 µm) from the experiment of Ho. Y.N [17] were used as the base case for analyzing these parameters further.
The moisture ratio curve and fluidization profile for particles with a size of 2000 µm were analyzed at air velocities of 0.6, 1.0, 1.3, 1.8, and 2.2 m/s and at a temperature of 50 °C. Increasing the air velocity shortens the required drying time to achieve the desired final moisture ratio. However, the effect of air velocity on the drying curve is less significant than that of the fluidization

Conclusions
A simulation was performed to examine the influence of inlet air velocity, inlet air temperature, and particle size on achieving the final moisture ratio of 0.25. The multiphase Eulerian model was used to describe the gas-solid turbulent flow. A 2D CFD model with a fine mesh was also developed. The small errors in the model validation indicated that this model is acceptable. The optimum parameters (i.e., 1.3 m/s, 50 • C, and 2000 µm) from the experiment of Ho. Y.N [17] were used as the base case for analyzing these parameters further.
The moisture ratio curve and fluidization profile for particles with a size of 2000 µm were analyzed at air velocities of 0.6, 1.0, 1.3, 1.8, and 2.2 m/s and at a temperature of 50 • C. Increasing the air velocity shortens the required drying time to achieve the desired final moisture ratio. However, the effect of air velocity on the drying curve is less significant than that of the fluidization profile. An air velocity of 1.0 m/s to 2.2 m/s was found to be suitable for the fluidization of particles with a size of 2000 µm.
The inlet air temperatures of 50, 60, 70, and 80 • C were manipulated at 1.3 m/s for those particles with a size of 2000 µm. The results indicated that the final moisture content can be achieved within a short drying time at a high air temperature. Such a temperature also results in a high heat transfer and drying rate.
Different fluidization patterns were observed for particles with sizes of 200, 500, 1000, and 2000 µm at a velocity of 1.3 m/s and temperature of 50 • C. A smaller particle size also leads to a greater bed expansion. However, such an expansion also depends on the air velocity. Those particles with a size range of 500 µm to 2000 µm were found to be able to undergo fluidization at 1.3 m/s. Some improvements need to be implemented to obtain better and more accurate results. Given that only a few studies have examined the drying of sago waste, further experiments must be conducted to determine the thermodynamic and thermal properties of these materials. A more complex CFD model that includes both the evaporation of sago moisture into inlet air and the mass and moisture transfer must also be employed in future studies. The interaction process may also require the utilization of a user-defined function code.  . m pq mass transfer from p th phase to q th phase; . m qp mass transfer from q th phase to p th phase. = τ q q th phase shear stress tensor; u q shear viscosity of q phase; λ q bulk viscosity of q phase; → F q external body forces; → F li f t,q lift forces; → F wl,q wall lubrication forces; → F td,q virtual mass and turbulent dispersion forces; → R pq interaction force between phases; p pressure shared by all phases. h q specific enthalpy of q th phase; → q q heat flux; S q source of enthalpy; Q pq intensity of heat exchange between p th and q th phases; h pq interphase enthalpy; G k generation of turbulence kinetic energy due to mean velocity gradients; G b generation of turbulence kinetic energy due to buoyancy; Y M contribution of the fluctuating dilatation in compressible turbulence to overall rate; C 1 , C 2 , and C 3 constants; S , S k user source term; MR moisture ratio; k drying constant; T air temperature; v air velocity; t drying time (mins); a constant = 2.0888.