This article is an openaccess article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).
Thermochemical conversion of biomass offers an efficient and economically process to provide gaseous, liquid and solid fuels and prepare chemicals derived from biomass. Computational fluid dynamic (CFD) modeling applications on biomass thermochemical processes help to optimize the design and operation of thermochemical reactors. Recent progression in numerical techniques and computing efficacy has advanced CFD as a widely used approach to provide efficient design solutions in industry. This paper introduces the fundamentals involved in developing a CFD solution. Mathematical equations governing the fluid flow, heat and mass transfer and chemical reactions in thermochemical systems are described and submodels for individual processes are presented. It provides a review of various applications of CFD in the biomass thermochemical process field.
The use of biomass as a CO_{2}neutral renewable fuel is becoming more important due to the decreasing resources of fossil fuel and their effect on global warming. Thermochemical conversion of biomass offers a possible process to provide gaseous, liquid and solid fuels and prepare chemicals derived from biomass. Many efforts have been done on making thermochemical processes more efficient and economically acceptable. A significant portion of these efforts over the past two decades has focused on the development of numerical models of thermochemical reactors (such as gasifiers, pyrolyzers, boilers, combustors, incinerators) that can help to design and analyze the thermochemical process. Due to a combination of increased computer efficacy and advanced numerical techniques, the numerical simulation techniques such as CFD became a reality and offer an effective means of quantifying the physical and chemical process in the biomass thermochemical reactors under various operating conditions within a virtual environment. The resulting accurate simulations can help to optimize the system design and operation and understand the dynamic process inside the reactors.
CFD modeling techniques are becoming widespread in the biomass thermochemical conversion area. Researchers have been using CFD to simulate and analyze the performance of thermochemical conversion equipment such as fluidized beds, fixed beds, combustion furnaces, firing boilers, rotating cones and rotary kilns. CFD programs predict not only fluid flow behavior, but also heat and mass transfer, chemical reactions (e.g. devolatilization, combustion), phase changes (e.g. vapor in drying, melting in slagging), and mechanical movement (e.g. rotating cone reactor). Compared to the experimental data, CFD model results are capable of predicting qualitative information and in many cases accurate quantitative information. CFD modeling has established itself as a powerful tool for the development of new ideas and technologies.
However, CFD modeling for biomass thermochemical conversion still face significant challenges due to the complexity of the biomass feedstock and the thermochemical process. Biomass is a mixture of hemicellulose, cellulose, lignin and minor amounts of other organics with proportion and chemical structure affected by variety. Inorganic ash is also part of the biomass composition. The complex structure makes biomass compositions pyrolyze or degrade at different rates by different mechanisms and affect each other during thermochemical process, and it makes the biomass particle feedstock has anisotropic properties in physical characterization [
In this paper, we attempt to summarize the current state of various CFD applications concerning the biomass thermochemical conversion process. The challenges faced by modelers using CFD in the biomass pyrolysis are also discussed.
Computational fluid dynamics is a design and analysis tool that uses computers to simulate fluid flow, heat and mass transfer, chemical reactions, solid and fluid interaction and other related phenomena. Comparing to the physical experiment operation, CFD modeling is cost saving, timely, safe and easy to scaleup. CFD codes turn computers into a virtual laboratory and perform the equivalent “numerical experiments” conveniently providing insight, foresight and return on investment. Various numerical techniques known as direct numerical simulation (DNS), vortex dynamics and discretization methods have been employed in the solution of the CFD model equations. The most widely used numerical techniques are discretization methods mainly including finite difference (usually based on Taylor’s series, polynomial expansions), finite elements (based on calculus of variations, and the methodofweightedresiduals) and finite volumes method (based on controlvolume formulation). Finite difference techniques are rarely used in engineering flows due to the difficulties in the handling of complex geometry [
Biomass thermochemical conversion refers to the processes of biomass gasification for gaseous fuel or syngas, fast pyrolysis for liquid biooil, carbonization for solid carbon or combustion for heat energy. The differences among these thermal processes are determined by the operation conditions of feed properties, oxidizer (air, oxygen or steam) amount, temperature, heating rate and residence time. These conditions change the proportions of the gas, liquid and solid products.
CFD models of the thermochemical processes include description of fluid flow, heat and mass transfer, and chemical reactions. The process fundamental governing equations are the conservation laws of mass, momentum, energy and species, namely the following
CFD enforces these conservation laws over a discretized flow domain in order to compute the systematic changes in mass, momentum and energy as fluid crosses the boundaries of each discrete region [
The biomass thermo conversion includes complex chemical and physical processes such as vaporization, devolatilization, volatile secondary reactions, char oxidation, coupled with the transport phenomena. Many studies have been made and many models have been built to describe the process [
The devolatilization process begins when the biomass temperature reaches a critical level. Many biomass devolatilization models have been developed and several reviews of these models have been made [
The onestep global mechanisms can be shown as:
The reaction kinetic rate (k) is expressed in singlestep Arrhenius fashion as
For twostep Arrhenius reaction schemes, the kinetic devolatilization rate expressions of the form proposed by Kobayashi [
The major limitation of onestep global schemes is that they are neither able to predict the composition of volatiles nor account for various components of the virgin biomass. Onestep multireaction schemes have been developed to address these shortcomings and can be shown as:
One of the more recent developments in onestep multireaction schemes for biomass fuels is the use of the distributed activation energy (DAE) approach.
The major shortcoming of the onestep multireaction schemes is that they neglect secondary reactions (cracking of tar to light molecular weight volatiles). Multistep semiglobal schemes attempt to address this shortcoming of multireaction schemes by considering reaction routes for both primary and secondary reactions. There are many literature positions which introduced the kinetics data of these mechanisms.
Another general biomass devolatilization model is developed extending the chemical percolation devolatilization (CPD) model from coal. The CPD model is extended to devolatilization of biomass major components based on the consideration of their chemical structure and its transformation under various mechanisms. The model considers multiple mechanisms, including bridge breaking and rearranging, sidechain cracking and gas release, tar distillation, and crosslinking. The same reaction scheme is applied for biomass as for coals:
The chemical structure parameters in the original CPD model are defined directly taken from ^{13}
The devolatilization tar is a mixture of condensable hydrocarbons. The secondary tar crack reactions occur homogeneously in the gas phase or heterogeneously at the surface of the biomass or char particles. Tar is a complex mixture of many kinds of components and the cracking mechanism is very comprehensive. In the present study, tar cracking is considered to follow the overall reaction schemes such as:
Many experimental investigation and model studies have been done on the cracking process. The model stoichiometric coefficients and kinetics data can be found in the literatures [
The biomass devolatilization and cracking gas species will react with the supplied oxidizer and with each other such as water gas shift reaction. The heat generated by exothermic reactions is important for the release of volatiles and ignition of char. The common homogeneous reactions are:
More reaction mechanisms and the kinetic parameters can be found from the literature [
Char is the solid devolatilization residue. Heterogeneous reactions of char with the gas species such as O_{2} and H_{2}O are complex processes that involve balancing the rate of mass diffusion of the oxidizing chemical species to the surface of biomass particle with the surface reaction of these species with the char. The overall rate of a char particle is determined by the oxygen diffusion to the particle surface and the rate of surface reaction, which depend on the temperature and composition of the gaseous environment and the size, porosity and temperature of the particle. The commonly simplified reactions models consider the following overall reactions:
The literature positions that introduced and reviewed the char surface reactions and the kinetic relationship can be found [
Although NavierStokes equations are viewed as the basis of fluid mechanics describing the conservation laws of mass, momentum, and energy, they have a limited amount of applications in the areas of biomass thermochemical conversion. The additional processes may influence the dynamics of the thermochemical reactor system. The basic governing equations need to be strengthened with special additional physical models or assumptions to fully represent the physical process. The important additional models include turbulence models, porous media and multiphase models, heat transfer with radiation models, and mass transfer and diffusion.
Turbulent flows are characterized by fluctuating velocity fields primarily due to the complex geometry and/or high flow rates. Turbulence affects the heat and mass transfer and plays an essential role in some processes such as biomass gasification/pyrolysis in fluidized bed and nonpremixed combustion in furnaces. The NavierStokes equations can be solved directly for laminar flows, but for turbulent flows the direct numerical simulation (DNS) with full solution of the transport equations at all length and time scales is too computationally expensive since the fluctuations can be of small scale and high frequency. The DNS is only restricted to simple turbulent flows with low to moderate Reynolds numbers. In the cases of high Reynolds number flows in complex geometries, a complete timedependent solution of the instantaneous NavierStokes equations is beyond the nowadays computational capabilities. Hence, turbulence models are required to account for the effects of turbulence rather than simulate it directly in practical engineering applications. Two alternative methods are employed to transform the NavierStokes equations so that the small eddies do not have to be directly simulated: Reynolds averaging and filtering. Both methods introduce additional terms in the governing equations that must be modeled for turbulence closure.
The Reynoldsaveraged NavierStokes (RANS) equations represent transport equations for the mean flow quantities only, with all the scales of turbulence being modeled. The RANS models are developed by dividing the instantaneous properties in the conservation equations into mean and fluctuating components, as shown as:
The Favreaveraging (densityweighted averaging) of the flow field variables is used to account for the effects of density fluctuations due to turbulence. The classical Reynolds averaging technique brings unclosed Reynolds stress terms in the timeaveraged conservation equations and need be modeled for turbulence closure. The Reynoldsaveraged approach is generally adopted for practical engineering calculations.
Most common RANS models employ the Boussinesq hypothesis (eddy viscosity concept, EDC) to model the Reynolds stresses terms. The hypothesis states that an increase in turbulence can be represented by an increase in effective fluid viscosity, and that the Reynolds stresses are proportional to the mean velocity gradients via this viscosity. Models based on this hypothesis include SpalartAllmaras, standard kε, RNG kε, Realizable kε, kω and its variants [
The Reynolds stress model (RSM) closes the Reynoldsaveraged NavierStokes equations by solving transport equations for the Reynolds stresses directly, together with an equation for the dissipation rate. The RSM accounts for the effects of streamline curvature, swirl, rotation, and rapid changes in strain rate in a more rigorous manner than oneequation and twoequation models. The fidelity of RSM predictions is still limited by the closure assumptions employed. The modeling of the pressurestrain and dissipationrate terms is particularly challenging, and often considered to be responsible for compromising the accuracy of RSM predictions. However, use of the RSM is a must when the flow features of interest are the result of anisotropy in the Reynolds stresses, for examples the cyclone flows or highly swirling flows in combustors.
Large eddy simulation (LES) solves “filtered” transport equations by permitting direct simulation of large scale turbulent eddies. Filtering removes eddies that are smaller than the filter size, which is usually taken as the mesh size. The filtering process creates additional unknown terms that must be modeled in order to achieve closure. LES provides an accurate solution to the large scale eddies akin to DNS while the smaller eddies below the filter size are modeled. This is because the large turbulent eddies are highly anisotropic and dependent on both the mean velocity gradients and the flow region geometries, while smaller eddies possess length scales determined by the fluid viscosity and are consequently isotropic at high Reynolds numbers. LES offers an alternative method of reducing the errors caused by RANS and providing a more accurate technique for turbulence simulation. However, application of LES to biomass industrial engineering is still in its infancy for it is computational expensive [
The radiative transfer equation (RTE) for an absorbing, emitting, and scattering medium at position
A semitransparent medium is considered and the refractive index is equal to unity. The optical thickness
The Discrete Ordinates Model (DOM) solves the radiative transfer equation (RTE) for a finite number of discrete solid angles, each associated with a vector direction
The standard form DOM suffers from a number of serious drawbacks, such as false scattering and ray effects. Perhaps the most serious drawback of the method is that it does not ensure conservation of radiative energy. This is a result of the fact that the standard discrete ordinates method uses simple quadrature for angular discretization. Thus, it is a logical step in the evolution of the method to move to a fully finite volume approach, in space as well as in direction. The finite volume method uses an exact integration to evaluate solid angle integrals and the method is fully conservative [
P1 model is the simplest formulation of the more general PN radiation model, which is based on the expansion of the radiation intensity I into an orthogonal series of spherical harmonics. The method of spherical harmonics provides a vehicle to obtain an approximate solution of arbitrary high order (i.e. accuracy), by transforming the radiative transfer equation into a set of simultaneous partial differential equations. Using only four terms in the series solution of the respective differential equation, the following relation is obtained for the radiation flux:
The Rosseland radiation model can be derived from the P1 radiation model with some approximations. The radiative heat flux vector in a gray medium is approximated by
The Rosseland radiation model differs from the P1 model in that the Rosseland model assumes the intensity equal to the blackbody intensity at the gas temperature. Thus,
This model is also called “diffusion approximation” model, since the radiation problem reduces to a simple conduction problem with strongly temperature dependent conductivity. It is important to keep in mind that the diffusion approximation is not valid near a boundary [
The main assumption of the Discrete Transfer Radiation Model (DTRM) is that the radiation leaving the surface element in a certain range of solid angles can be approximated by a single ray. The equation for the change of radiant intensity,
Here, the refractive index is assumed to be unity. The DTRM integrates
The “ray tracing” technique used in the DTRM can provide a prediction of radiative heat transfer between surfaces without explicit viewfactor calculations. The accuracy of the model is limited mainly by the number of rays traced and the computational grid [
The mixture fraction model is used to present the reaction chemistry in the probability density function (PDF) method for solving turbulentchemistry interaction. The equilibrium model is applied which assumes that the chemistry is rapid enough for chemical equilibrium to always exist at the molecular level. Basing on the simplifying assumptions, the instantaneous thermo chemical state of the fluid is related to the mixture fraction
Under the assumption of equal diffusivities, the species equations can be reduced to a single equation for the mean (timeaveraged) mixture fraction f̄. And the mean mixture fraction variance
The source term
The porous media assumption is generally used in the applications of biomass pyrolysis in fixed bed. The arrangement of biomass particles in the fixed bed forms void spaces. The devolatilization volatiles and gases through the particle voids can be described as flow through a porous media. The particle position may change during the conversion process for the devolatilization, combustion and shrinkage of biomass particles. In this process to mesh all associated geometry with a complex unstructured or body fitted system is out of both computational power and CFD algorithms levels. Therefore, the simplified porous media assumption applies Darcy’s law to present the relationship on pressure drop and volume averaged velocity caused by viscous drag:
At high flow velocities, the modification of this law provides the correction for inertial losses in the porous medium by DarcyForchemier equation:
Fluid flow, and heat and mass transfer are described in the subdomain by the laws of conservation of mass, momentum and energy in the terms of macroscopic variables provided by the volumeaveraged NavierStocks equations in a version of Darcy’s law. The system can be regarded as a twophase flow [
The flow in biomass fluidized bed gasifier or boilers and furnaces is a typical kind of gassolid flow with chemical reactions. Thus hydrodynamics of the gassolid flow can be performed based on the Eulerian–Lagrangian concept. The discrete phase method can be applied to the particle flow when the particle phase can be considered to be sufficiently dilute that the particleparticle interactions and the effects of the particle volume fraction on the gas phase can be assumed neglected. The coupling of the continuous phase and the discrete phase is important and it is solved by tracking the exchange of mass, momentum and energy.
The model computes the particle trajectory using a Lagrangian formulation which includes the inertia, hydrodynamic drag, and the force of gravity. The particle trajectory can be predicted for the
Biomass gasification and pyrolysis are thermally degraded processes in insufficiency or absence of air/oxygen aiming at the production of solid (charcoal), liquid (tar/biooil) and gaseous products. The CFD models used to describe these processes have become an important analysis and design tool to achieve the flow and temperature pattern, the products concentration contour and yields.
Fletcher
Gerun
The largest application of CFD models has been to power station boilers and furnaces. Many studies made in relation to coal combustion have been modified to apply to biomass combustion or cofiring.
Dixon
Kær
The cofiring of coal and biomass has been advocated for a number of years as being advantageous on both an environmental and economic basis. The cocombustion of biomass as a minor component presents an interesting intermediate situation with a high reactivity solid. There are a number of commercially available CFD models, and the suitability of the submodels available for biomass combustion is a key factor in selecting an appropriate code.
In the case of biomass burner studies there is considerable interest in
This paper summarized the CFD applications in biomass thermochemical conversion and system design. There is evident that CFD can be used as a powerful tool to predict biomass thermochemical processes as well as to design thermochemical reactors. CFD has played an active part in system design including analysis the distribution of products, flow, temperature, ash deposit and
Twostage semiglobal reaction schemes for: (a) cellulose; (b) wood.
The geometry of the gasifier. The lower inlets are used to inject the biomass mixed with air, and the upper inlets are used to inject steam [
Temperature profile in the reactor [
Velocity pattern in the reactor [
Flow simulations for the asconstructed design: (a) Gas velocity; (b) particle trajectory [
Flow simulations for the modified design: (a) Gas velocity; (b) particle trajectory [
Predicted deposition mass flux in gm^{−2}h^{−1} [
Closeup of the secondary super heater showing boundary layer controlled deposition flux in gm^{−2}h^{−1}. [
Closeup of the secondary super heater showing vapour deposition flux in gm^{−2}h^{−1}. [
Predicted particle traces coloured by particle mass (kg) for Thoresby coal–biomass combustion cases: 0.75 mm diameter biomass particles [
Predicted contours of potassium concentration (mol/mol) [
Predicted NO formation in the furnace through the NH3 route (mol/mol) [
Thermochemical conversion variant
Technology  Residence time  Heating rate  Temperature °C  Aim Products  Oxidizer amount 

carbonation  very long (days)  low  low (~400)  charcoal  absence 
fast pyrolysis  short (<2 sec)  high (>1000°C/s)  moderate (~500)  biooil, chemicals  limited 
gasification  long  high  high (~800)  Gas, chemicals  limited 
combustion  long  high  high  heat  enough 
CFD applications in biomass gasification and pyrolysis.
Application  Code  Dim  Aim/Outcome  Turb. Model  Extra Model  Agreement with Exp.  Authors 

Entrained flow gasifier [ 
CFX4  3D  Products mass fraction distribution; temperature contours; swirl velocity distribution  Std

Lagrangia n  Acceptable  Fletcher, D. F. 
Twostage downdraft gasifier [ 
Fluent  2D  To investigate in detail the oxidation zone; temperature profile; velocity pattern; tar conversion mechanism study  RNG

DOM  Satisfactory  Gerun, L. 
Horizontal entrainedflow reactor [ 
Fluent  2D  Predictions of flow, temperature and conversion; sensitivity of the kinetic parameters of pulverized corn stalk fast pyrolysis  n/a  Lagrangia n  Reasonable  Xiu, S. N. 
Cone calorimeter reactor [ 
Code  3D  To model heat transfer and pyrolysis within dry and wet wood specimens, and the mixing and pilot ignition of the released volatiles  n/a  Porous  n/a  Yuen, R. K. K. 
Moving packed bed [ 
Fluent  2D  Detailed comparisons between the combustion mode and gasification mode in a waste movinggrate furnace  Std

DOM  n/a  Yang, Y. B. 
Entrained flow gasifier [ 
CFX  2D  To model black liquor gasification, model parameters identification and sensitivity analysis  Std

Lagrangia n

n/a  Marklund , M. 
Downdraft gasifier [ 
Code  3D  Temperature profile, pressure drop, model parametric analysis  n/a  Porous  n/a  Sharma, A. K. 
Fluidized bed flash pyrolysis [ 
Code  3D  An integrated model proposed to predict wood fast pyrolysis for biooil  n/a  Radiation  Good  Luo, Z. Y. 
Dim=Dimension, Turb=Turbulence, Std=Standard, DOM = Discrete Ordinates Model (radiation), DTRM=Discrete Transfer Radiation Model, exp=experiment,
CFD applications in biomass combustion.
Application  Code  Dim  Aim/Outcome  Turb. Model  Extra Model  Agreement with Exp.  Authors 

Bagasse fired boilers[ 
Furnace  3D  Tube erosion; heat transfer Airheater corrosion; Swirl burner  Std

Lagrangian; porous  Acceptable  Dixon, T. F. 
Strawfired grate boiler [ 
CFX  3D  To provide insight into the boilers; heat transfer predictions; To predict ash deposition  RNG

DTRM  Good  Kær, S. K. 
Combustion Furnace[ 
Fluent  3D  Particle tracks, temperature contours  Std

Lagrangian; DOM  n/a  Shanmukharadhya, K. S. 
Waste rotary kiln incinerator [ 
Fluent  3D  To describe the processes occurring within the gaseous phase of the kiln and of the post combustion chamber  Std

P1  n/a  Marias, F. 
Bagassefired furnaces [ 
Fluent  3D  To gain insight into the effect of moisture on the flame front.  Lagrangian; P1  n/a  Shanmukharadhya, K. S.  
Tube stove[ 
CFXTASCf low  3D  To understand the aerothermochemical behaviour of the stove operation in combustion and gasification modes  n/a  cphase  Excellent  Dixit, C. S. B 
Wastetoenergy plant[ 
Fluent

To maximize the energy recovery efficiency of wastetoenergy plants  DOM  n/a  Goddar, C. D. 
Dim=Dimension, Turb=Turbulence, Std=Standard, DOM = Discrete Ordinates Model (radiation), DTRM=Discrete Transfer Radiation Model , P1=P1 radiation model, exp=experiment
CFD applications in biomass cofiring.
Application  Code  Dim  Aim/Outcome  Turb. Model  Extra model  Agreement with Exp.  Authors 

Biomass and coal cofired[ 
CINAR  3D  A new approach based on neural networks is proposed  Radiation; Lagrangian  n/a  Abbas, T.  
Cofiring[ 
Fluent 6.1  3D  To predict the behaviour of the biomass in the coal flame.  RNG

P1

n/a  Backreedy, R. I. 
Cofiring combustors [ 
Fluent UDF code  To develop a fragmentation subroutine applicable to Fluent via a UDF.  n/a  Lagrangian; fragmentation model  Reasonable  Syred, N.  
Cocombustion boilers[ 
Fluent 6.1 MAT LAB  3D  To optimize burner operation in conventional pulverizedcoalfired boilers  Std 
DOM  n/a  Tan, C. K. 
Biomass utility boiler[ 
Fluent 5.6  3D  To examine the impact of the large aspect ratio of biomass particles on carbon burnout in cofiring switchgrass/coal.  Std 
Lagrangian; DOM  n/a  Gera, D. 
Dim=Dimension, Turb=Turbulence, Std=Standard, DOM = Discrete Ordinates Model (radiation); P1=P1 radiation model, exp=experiment
CFD applications in
Application  Code  Dim  Aim/Outcome  Turb, Model  Extra Model  Agreement with Exp.  Authors 

Test furnace [ 
Code  3D  Particle tracks, temperature contours, NO formation, potassium concentration  RNG

Lagrangian; P1; radiation; 
Good  Ma, L. 
Combustion chamber [ 
Fluent 5.5  3D  Prediction of gaseous emission  SST 
Lagrangian; DTRM; 
Good  Miltner, M. 
Pilot downfired combustor [ 
Fluent 5.0  3D  To describe the processes occurring within the gaseous phase of the kiln and of the post combustion chamber  P1; Lagrangian; 
n/a  Zarnescu, V.  
Fluidized beds [ 
Fluent 6.2  3D  To compare the performance of five global ammonia chemistry mechanisms in fullscale boiler CFD modeling.  Std

DOM; Global Ammonia Chemistry Mechanism s  Well under special conditions  Saario, A. 
Biomass combustion [ 
Code  1D  Comparisons of the Validity of Different Simplified NH3Oxidation Mechanisms for Combustion of Biomass  n/a  Ammonia oxidation mechanisms  n/a  Norstrom, T. 
Wood stove [ 
Spider  2D  To model nitricoxide formation from fuelbound nitrogen in biomass turbulent nonpremixed flames.  Std

DTRM  n/a  Weydahl, T. 
Bagassefired boiler [ 
Furnace  3D  To apply conditional moment closure (CMC) in a to obtain predictions of CO and NO in the flue gas.  Std

Lagrangian; DTRM; PDF; conditional moment closure equation  Reasonable  Rogerson, J. W. 
Dim=Dimension, Turb=Turbulence, Std=Standard, DOM = Discrete Ordinates Model (radiation); DTRM=Discrete Transfer Radiation Model, P1=P1 radiation model, PDF= Probability Density Function, exp=experiment.