Development and Validation

18th Symposium on Thermal Science and Engineering of Serbia Sokobanja, Serbia, October 17–20, 2017 Society of Thermal Engineers of Serbia

Faculty of Mechanical Engineering in Niš

Mathematical Model of Pulverized Coal Devolatilization and Combustion in a Swirl Burner– Development and Validation Aleksandar R. Milićevića(CA), Srđan V. Beloševića, Ivan D. Tomanovića,Nenad Đ. Crnomarkovića and Dragan R. Tucakovićb University of Belgrade, Vinča Institute of Nuclear Sciences, Belgrade, RS, [email protected] University of Belgrade, Vinča Institute of Nuclear Sciences, Belgrade, RS, [email protected] a University of Belgrade, Vinča Institute of Nuclear Sciences, Belgrade, RS, [email protected] a University of Belgrade, Vinča Institute of Nuclear Sciences, Belgrade, RS, [email protected] b University of Belgrade, Faculty of Mechanical Engineering, Belgrade, RS, [email protected] a

a

Abstract:In this paper a new comprehensive mathematical model of pulverized coal combustion for prediction of turbulent transport processes and reactionswasdeveloped and validated. Numerical simulations were carried out by using an in-house developed computer code, with sub-models for individual phases during complex combustion process: evaporation, devolatilization, combustion of volatiles and char combustion. The model is based on Euler-Lagrangian approach to a two-phaseflow with standard k-ε turbulence model, PSI-CELLmethod for taking into account mutual influences of phases,diffusion model for particle dispersionand four-fluxes radiation model.For sub-model of coal devolatilization the single rate kinetic model with the approach of Merrick is adopted, for volatiles combustion the finite rate/eddy break up model is chosen, while for char oxidation the combined kineticdiffusion model is used.The obtained numerical results are quite satisfactory in comparison with available measured data from a 150 kW experimental swirl burner combustion chamber,especially with respect to complexity of the considered problem.It is concluded that the developed model, with necessary modifications, can be successfully used for prediction and computation of parameters of swirl burners for a range of pulverized fuels, such as coal and biomass. Keywords:Combustion, Devolatilization, Mathematical modeling, Pulverized coal, Swirl burner.

1. Introduction The two-phase multicomponent turbulent flowswith processesof fuel combustion are highly complex, with many mutually coupled impacts. CFD optimizationof these processesunder a wide range of operating conditions can produce large savings in equipment and energy costsand in reduction of environmentalpollution [1].The sub-models of pulverized coal combustion contain mathematical and numerical representations of the fundamental principles that characterize the physicochemical phenomena of interest [2]. In this paper a new comprehensive mathematical model of pulverized coal combustionfor prediction of turbulent transport processes with special focus on chemical reactionswasdeveloped and validated against available measurements. Devolatilization process plays a significant role in combustion pulverized coal and two main approaches are used for devolatilization modeling: network devolatilization models and empirical devolatilization models [3]. The network models are able to accurately predict the rates, yield and composition of volatile using as input experimental data, but they are quite-complex and require significantly larger computational resources in compare with empirical devolatilization models. In this work, the single kinetic rate model is chosen for process of devolatilization, but the empirical approaches, like this one, do not predict the composition of volatiles. Therefore it is necessary to define the sub-model for determination of the composition of volatiles, as well. The swirl burners used in pulverized coalare designed to ensure flame stability and to make short, intenseflames in an effort to reduce furnace size and increase fuel burnout. Both objectivesare met by swirling the secondary air prior to entering the furnace [4, 5]. The circulation of hotcombustion gases ensure flame stability, while swirled secondary air alsoincreases fuel and air mixing, producing short, intense flames. The main aim of this work is development of mathematical model and computer code representing

complex tool for analysis of processes during combustion, which would include pulverized solid fuels in a wider range of characteristics. The developed model has to be efficient for practical application and to enable the complex parametric analysis in real conditions.

2. Furnace geometry and firingparameters The model was validated with available experimental data in a 150 kW swirl-stabilized down-fired dual-feed burner flow reactor.The referent measurements were performed by Brigham Young University in United States [4].The reactor includes six separate sections stacked on top of each other, and each section includes four large (12 by 30 cm)removable windows that allow access to the reactor chamber. The down-scaled dual-feed swirl burneris placed on the top of the furnace and fires the coal downward into the reactor. The coaland transport (primary) air arefed into the burner through the center orifice, theannular orifice is empty in this case, while the outer orifice is used for theintroduction of secondary air. The schematic drawing of the dual-feed burner and furnaceinterior is presented graphically in Figure 1.

Figure 1. Dual-feed burner cross-section and scaled view furnace interior

The fuel used for combustion is Blind Canyon coal(a high volatile bituminous coal). The main properties of coal and firing parameters are given in Table 1. Table 1. Main properties of coal and firing parameters Proximate analysis data (the base: AR= as received; D=dry; DAF=dry ash-free; % on mass) Coal

Moisture [%, AR]

Volatiles [%, D]

Fixed carbon [%, D]

Ash [%, D]

2.1

40.6

51.5

7.89

Ultimate analysis data (% on mass) Coal

C [%, DAF]

H [%, DAF]

O [%, DAF]

N [%, DAF]

S [%, DAF]

81.22

5.52

10.97

1.66

0.63

Operational parameters of coalfiring Central fuel (coal) [kg h-1]

Annular fuel [kg h-1]

Central air [kg h-1]

Annular air [kg h-1]

Secondary air [kg h-1]

Swirl number

15.3

0

11

0

160

1

Particle size and coal particle density Mean diameter d p [μm] Coal

Density ρ p [kg m-3]

110.4

1400

3. Mathematical model of pulverized coal reactions and computer code In-house developed comprehensive combustion model in cylindrical coordinatesat stationary conditions is used for numerical predictionsof processes in pulverized coal-fired furnace. Two-phase flow is modeled by Eulerian–Lagrangian approach.Gas phase is described by time-averaged Eulerian conservation equations for mass, momentum, energy, gas mixture component concentrations, turbulence kinetic energy and its rate of dissipation, given for general variable Φ in index notation by using the following eq. (1):

∂ ∂  ∂Φ  Φ ρU j Φ=  Γ Φ  + SΦ + S p . ( ) ∂x j ∂x j  ∂x j 

(1)

Φ In previous equation S p is additional source due to particles, while ρ , U j , Γ Φ and SΦ represent gas-

phase: density [kg m-3], velocity components [m s-1], transport coefficient and source term for general variable Φ . The standard k-εturbulence model is used for closure of the gas phase momentum conservation equations. Radiative heat exchange in furnace is simulated by discrete ordinates model.Continuity equation for particle number density is given in the form of eq. (1).Particle-to-gas impact is considered by PSI Cell (Particle Source in Cell) method, whileparticle collisions are neglected. The motion of dispersed phase is modeled using Basset equation in a Lagrangian field, taking into account that drag and gravity forces acting on the particles [6], with less important effects neglected, as it shown in eq. (2):

 dup 1 d 3pπ      (2) − − + m= C ρ A u u u u ρp − ρ ) g . ( D p p p ( p) dt 2 6    -1 The particle velocity vector up [m s ] is a sum of convective upc and diffusion velocity upd .Convective  velocity upc is obtained by solving particle motion eq. (3):  dupc 1 d 3pπ      (3) ρp − ρ ) g . mp= CD ρ Ap u − upc ( u − upc ) + ( dt 2 6  The diffusion velocity upd is obtainedby using the eq. (4):

 1 upd = − Γ p ∇N p , Np

(4)

whereby Γ p [m2 s-1] and N p [m2 s-1]represent coefficient of turbulent diffusion of particles and the particle number densityin given control volume, respectively. Initial and boundary conditions usual for elliptical partial differential equations are applied. Conditions near the walls are described by the wall functions, which have significant impact on the accuracy of numerical results [7]. Moisture present in the coal is released during heating of the particles to about 100 °C and the rate of moisture evaporation is calculated by the kinetic rate expression (5):

dmp dt

= − A exp ( − E / RT ) mp .

(5)

In the previous equation A, E and R denote the frequencyfactor, activation energy and universal gas

A 5.13 × 1010 [s-1], = E 8.8 × 107 [J kmol-1], and R = 8.314 [J/ (mol·K)]. constant, respectively:= Further heating over 400-500 °C, the volatiles are released (combustible volatile matter). For process of devolatilization empirically single kinetic rate model (SRM) is selected, supported by the sub-model for determination of the composition and the amount of volatile. The SRM assumes that the rate of devolatilization is first-order dependent on the amount of volatiles remaining in the particle (6):

dmp = k vol  mp − (1 − f v ,0 )(1 − f w,0 ) mp,0  . dt

(6)

3.12 × 10 [s-1] and Evol = 7.4 × 10 [Jkmol-1]. Kinetic rate parameters of devolatilization coal k vol [s-1]are: A= vol The SRM does not predict the composition of the volatiles, and therefore the sub-models for determining the composition of volatiles of coal are defined, as well. A sub-model for the determination of the composition and amount of the coal volatiles is selected by using a matrix defined by Merrick in [8]. The sub-model determines the composition and amount of the volatiles, based on the known proximate and ultimate analysis of the fuel. For the prediction of effective volatile matter in function of the volatile matter proximate analysis, the expression (7) is used: 5

V= p − 0,36 p 2 .

7

(7)

The matrix proposed by Merrick makes a system of ten equations with ten unknowns, which is solved simultaneously. By introducing mathematical correction matrix, the five unknowns can be solved explicitly:CHAR=1-V, CH 4 =1.31ˑH, C 2 H 6 =0.22ˑH, CO=0.32ˑO and CO 2 =0.15ˑO. The remaining components of the coal volatiles: TAR, H 2 , H 2 O, NH 3 , and H 2 S, can be simultaneously solved by an implicit system of five equations with five unknowns (8):

 TAR  1 0,1111 0,1765 0, 0588  H 2  0 0,8889 0 0   H 2O  =    0 0 0,8235 0   NH 3  0 0 0 0,9412   H 2S  . C - 0,98 ⋅ CHAR - 0, 75 ⋅ CH 4 − 0,8 ⋅ C2 H 6 − 0, 4286 ⋅ CO − 0, 2727 ⋅ CO 2    H - 0, 002 ⋅ CHAR - 0, 25 ⋅ CH 4 − 0, 2 ⋅ C2 H 6    = O - 0, 002 ⋅ CHAR - 0,5714 ⋅ CO − 0, 7273 ⋅ CO 2   N - 0, 01 ⋅ CHAR     S - 0, 006 ⋅ CHAR  0,85 0, 082  0, 049  0, 009  0, 01

0

0

0

0

(8)

Products of the devolatilization of the pulverized coal are considered to contain: tar, the primary gaseous volatiles, and residual char. The tar further decomposes to secondary gaseous volatiles: CH 4 , HCN, H 2 , CO and residual soot. The elemental composition of tar is chosen in the same manner as in Bradley’s work [10] for laminar pulverized coal–air combustion. The residual soot in tar and carbon in coal char are oxidized directly, with ash remaining after oxidation of the char. The devolatilization model adopted is shown in Figure 2. Combustion of volatiles, which represents homogeneous reaction, is modeled according to equations (9), (10) and (11):

CO + 0,5O 2 → CO 2 ,

(9)

CH 4 + 2O 2 → CO 2 + 2H 2 O ,

(10)

H 2 + 0,5O 2 → H 2 O ,

(11)

with solving the conservation equations for components of the gas mixture: N 2 , O 2 , CO 2 , H 2 O, CH 4 , CO, H 2 and NH 3 . For turbulence chemistry interaction the finite rate/eddy break up model is adopted, in which processes are controlled by slower of the two processes: chemical kinetics and turbulent mixing [10]. Chemical kinetics is given by Arrhenius expression (12): b Ωch = Ah ⋅ xfua ⋅ xox ⋅ ρ c ⋅ exp ( − Eh / RT ) .

(12)

The turbulent mixing is determined according to the eddy-break up model by using the following eq. (13): x ε ε  = Ωct Afu min  ρ ⋅ xfu ⋅ , ρ ⋅ ox ⋅  . k s k 

(13)

Figure 2. Devolatilization model and product composition of coal

For char combustion, which represents the heterogeneous reaction, the kinetic/diffusion rate model has been selecting [10] within a “shrinking core” concept, as follows (14): Rc =

ox Ap ⋅ M p ⋅ xmol . 1 1 + k r kd

(14)

Also, oxidation of carbon from char is considered directly, as shown in the following eq. (15):

C + O 2 → CO 2 .

(15)

The values of kinetic parameters for homogeneous and heterogeneous oxidation reactions, used in previous equations, are presented in[11]. Discretization of partial differential equations is performed by control volume method andhybriddifferencing scheme. Discretized equations are solved by SIPSOL method. Coupling of continuity and momentum equations is done by SIMPLE algorithm. Stabilization of iteration procedure is provided by underrelaxation. The comprehensive combustion code was verified by grid-independence study.Numerical simulation is carried out by using the in-house developed CFD code, inwhich sub-models for coal particle phase during complex combustion process (evaporation, devolatilization, combustion of volatile, and char combustion) are implemented.

4. Numerical simulation For numerical simulation the structured numerical mesh has been selected, which consist of 19656cells ( 252 ( x ) × 78 ( r ) ) , on the basis of the grid independence study. Numerical run takes 5000 iterations in order to achieve the converged solutions. By using a computer program, describedpreviously, discrete values in certain numerical nodes of gridare obtainedfor the following variables: mean velocities of gas in axial, radial and tangential direction; pressure and temperature of gas; turbulence kinetic energy and turbulent kinetic energy dissipation; componentconcentrations in mixture; and concentration of coal particles. Particle trajectories are also calculated, that is coordinates of consecutive positions of each of them, and at this points the values of gas and particle velocity as well as particle and gas temperature. The well-converged results obtained by numerical simulations are compared with the experimental data in the reactor. Regarding the behavior of the flame, an internal and an external recirculation zone are formed in the top section of the furnace, just as expected for a confined swirl burner, as shown in Figure3. The internal recirculation zone forces gas flow toward the burner along a portion of the centerline and thereby slowing the entering fuel particles. An external recirculation zone is formed in the upper corners where the furnace wall meets the top, transporting hot combustion products to the top wall of the furnace.

Figure 3. Sketch of internal and external recirculation zone gas flow

The results show the physicallyexpected shape of the coal particle trajectory. The particles due to initial inertia penetrate in internal recirculation zone, where they are gradually stopped under influence of the reverse flow. Due to decreased axial velocity, radial and tangential components of velocity are beginning to prevail, so that the particles change direction and move towards the periphery of the chamber under the influence of vortex. Then, coal particles are abstracted by gas flow, which directs them to the wall of reactor; along which particles continue to move, and from which occasionally refuse, as shown in Figure 4.

Figure 4. Trajectoriesof pulverized coal particles

Gas sampling system was used in investigation [4] for producing quantitative maps of gas species in firing flame, which was serve as a benchmark for CFD validation. The most important components of the gas sampling system in discussing firing flame issues are the gas sampling probe with FTIR (Fourier transform infrared) and a Horiba gas analyzer. Also, the thesis [4] explained difficulties of measuring gas species and why the obtained experimental data did not exhibit perfect symmetry about the geometric centerline. Results of numerical simulation of CO 2 and O 2 are compared with the experimental data from [4] along the radial direction at different axial distancesdownstream of the burner for the combustion coal flame, as shown in Figure5. and Figure6.Given concentrations of combustion products reflect the complex turbulent transport

processes and reactions in the reactor.Results are shown in the entire cross-section of the furnace because the measured profiles are somewhat different due to the influence of the measuring device.

Figure 5. Comparasion of experimental data and CFD model results for molar fraction of O 2

Figure 6. Comparasion of experimental data and CFD model results for molar fraction of CO 2

The comparison between the predicted values of O 2 and CO 2 are found to show a quite satisfactory agreement with the measured data, which also agree with expectations on the profiles of the species. Towards the furnace outlet, the flow develops; the deviations become smaller and near the outlet of the reactor average discrepancies do not exceed 5%.

After they enter the reactor, due to the intensive heat exchange, pulverized coal particles undergo a series of transformations: moisture release, devolatilization, combustionof volatile compoundsand char residue combustion. The CFD results of temperature field in the furnace are shown in Figure7.Temperature was not measured in [4], so a qualitative analysis with obtained results was performed, and considering whether the model is physically and chemically real shows processes that are modeled. Based upon the temperature field, two zones with the highest temperature can be noticed: the firstat the entrance of the burner, where the volatiles startto combust;and the second, a little further from the entrance towards the wall, which isconsequenceofbeginning of the charcombustion.

Figure 7. Temperature profile of CFD simulations

Qualitatively, all profiles correspond to the reference measurements. According to the results of comparisons it could be concluded that the proposed mathematical model was successfully developed and validated.

5. Conclusions A new comprehensive differential mathematical model and computer code are developed and validated for investigationof coal combustionprocesses in a 150 kW experimental swirl burner combustion chamber. Numerical simulationsare carried out by an in-house developed computer program, which has incorporated solid particles tracking and combustion reactions sub-models for complex in-furnace process. The obtained numerical results are quite satisfactory in comparison with available measured data from a 150 kW experimental swirl burner combustion chamber,especially with respect to complexity of the considered problem. It is concluded that the developed model, with necessary modifications, can be successfully used for prediction and computation of parameters of swirl burners for a range of pulverized fuels, such as coal and biomass. Suchmathematical modelof pulverized coal combustion for prediction of turbulent transport processes and reactions,could be a basic for complex numerical analysis, with numerous parameters varied (fuel and air distribution, fuels particles shape and size…)in order to optimize the processes during combustion, regarding both the emission and the efficiency.

Nomenclature Latin symbols

f w,0 – Mass fraction of moisture.

A, Ah – Frequency factors in Arrhenius relations, in [s-1].

k

Afu

– Coefficient for homogeneous reaction.

Ap

– Particle cross section area, in [m2].

f v ,0 – Mass fraction of volatiles initially present in the particle.

– Turbulence kinetic energy, in [m2s-2].

kr – Reaction rate parameter in kinetic regime, in [s-1].

kd [s-1].

– Diffusion parameter of mass transfer, in

Mp

– Molar mass of particle, in [kgmol-1].

mp

– Particle mass, in [kg].

ρ

mp,0

– Initial particle mass, in [kg].

Ωch , Ωct – Kinetic and turbulent mixing reaction

p

– Proximate analysis of volatile matter.

s

– Stoichiometric coefficient.

rates of homogeneous reaction, in[kgm-3s-1].

T

– Temperature, in [K].

t

– Time, in [s].

V

– Effective volatile matter.

Rc

– Heterogeneous reaction rate, in [kgs-1].

Superscripts andSubscripts a , b, c

– Coefficients of homogeneous reaction

fu – Fuel mol

– Molar

ox

– Oxidant

p

– Particle

t

– Time interval

0

– Environment

and oxidant, in [kgkg-1].

v,vol

– Volatile

– Oxidant molar concentration,[molm-3].

w

– Wet

Greek symbols – Turbulent kinetic energy dissipation, in ε 2 -3 [m s ]. xfu , xox – Mass concentrations of combustible gas ox xmol

– Fluid density, in [kgm-3].

Acknowledgement This work was supported by the Ministry of Education, Science and Technological Development of the Republic of Serbia (project: “Increase in energy and ecology efficiency of processes in pulverized coal-fired furnaces and optimization of utility steam boiler air preheater by using in-house developed software tools”, No. TR-33018).

References [1]

Ferziger, J., Peric, M., Computational Methods for Fluid dynamics, Springer-Verslag Berlin Heidelberg New York, 2002 [2] Belošević,V. S., Tomanović,D. I., Crnomarković,D. N.,Milićević,R. A., Tucaković,R. D.,Numerical Study of Pulverized Coal-fired Utility Boiler over a wide Range of Operating Conditions for in-Furnace SO 2 /NO x Reduction, Applied Thermal Engineering,94 (2016),pp. 657–669 [3] Smith, L., Douglass,S., Fletcher,T., The Structure and Reaction Processes of Coal, Plenum Press, New York, USA, 1994 [4] Damstedt, B.D., Structure and Nitrogen Chemistry in Coal, Biomass, and Cofiring Low-NO x Flames, Ph.D. thesis,Brigham Young University, Provo, Utah, 2007 [5] Saljnikov A., Sagorevanje ugljenog praha u osnosimetričnom turbulentnom strujnom polju, Ph.D. thesis,University of Belgrade – Faculty of mechanical engineering, Belgrade, Serbia, 1998 [6] Sijerčić, M., Matematičko modeliranje kompleksnih turbulentnih transportnih procesa(Mathematical Modeling of Complex Turbulent Transport Processes, in Serbian), Yugoslav Association of Thermal Engineers and VINČA Institute of Nuclear Sciences, Belgrade, 1998 [7] Milićević, A., Tomanović,I., Stojanovic, A., Belošević,S., Stankovic. B.,Crnomarković,N., Sijercic. M., Wall Functions in the Thermal Energy Equation within k-ε Turbulence Model, Proceedings, International Symposium Power Plants2014, Zlatibor, Serbia, October 28-31, 2014, pp. 719–728 [8] Merrick, D., Mathematical Models of the Thermal Decomposition of Coal: 1. The Evolution of Volatile Matter, Fuel, 62 (1983), 5, pp. 534–539 [9] Bradley, D., Lawes M.,Park, H-Y., Usta, N., Modeling of Laminar Pulverized Coal Flameswith Speciated Devolatilizationand Comparisons with Experiments, Combustion and Flame, 144 (2006), 1-2, pp. 190–204 [10] Belošević, S., Sijerčić,M., Stefanović,P., A Numerical Study of Pulverized Coal Ignition by Means of Plasma Torches in Air–Coal Dust Mixture Ducts of Utility Boiler Furnaces, International Journal of Heat and Mass Transfer, 51 (2008), 7-8, pp. 1970–1978 [11] Yin, C., Rosendahl,L., Kær,S., Towards a Better Understanding of Biomass Suspension Co-firing Impacts via Investigating a Coal Flame and a Biomass Flame in a Swirl-Stabilized Burner Flow Reactor under Same Conditions, Fuel Processing Technology, 98 (2012), pp. 65–73



Modeling for pyrolysis of wood particles Biljana Miljković Faculty of Technical Sciences, Novi Sad, Serbia, [email protected] Abstract: Compared with coal, biomass is characterized by a higher content of volatile matter. It is a renewable source of energy and has many advantages from an ecological point of view. Understanding the physical phenomena of pyrolysis and representing them with an mathematical model is basic step in design of pyrolysis reactors. In this study, an original numerical model for the pyrolysis of wood chips is proposed and relevant equations solved using original program realized in MATLAB. To check the validity of the numerical results, woody biomass was used as experimental sample in laboratory facility. Keywords: mathematical modeling, pyrolysis, experiment, biomass

1. Introduction Biomass, especially wood, is one of the first sources of energy used by humans. In developing countries it is still the major source of energy, while in developed countries a renewed interest in biomass started in the nineteen-seventies. In Serbia, the major energy demand is provided from fossil fuels. Currently, it is not possible to satisfy energy needs with available domestic primary energy sources and therefore 30-40% of annual energy needs are imported (5.31 Mtoe in 2010. of oil)[5]. On the other hand, the total renewable energy potential in Serbia is more than 4.3 Mtoe per year [6]. Unfortunately, the production of energy from renewable sources in Serbia is still present only in some small plants. Biomass is not currently used for electricity generation, but just new facilities are installed in food and processing industry. In spite of that, the production of energy from renewable sources is just a significant potential for Serbia at this moment. The most important source of biomass in Serbia is in agriculture and forestry (2.7 Mtoe) [6]. Generally, the northern part of the territory of Serbia is plain agricultural area, while the southern part is a mountainous region rich in forests. It is clear that biomass cannot totally replace fossil fuels. However it may be a partial answer to the problems of CO2 emissions and oil dependency. It appers that biomass, especialy wood, is a much cleaner fuel than coal. Wood has a very low sulfur content. There is usually no need for desulfurisation tretment of the flue-gas in wood combustion. The combustion temperature is lower due to a lower HHV, which reduces the fuel and thermal NO x formation. However, a denitrification installation might still be necessary. On the other hand, energy density for coal is 3 to 5 times greater than for wood. Hence, for wood to be cost competitive, it is important to limit fuel transportation and storage needs. For that reason, wood power plants are usually located in forested areas. The thermo-chemical conversion (pyrolysis, gasification, combustion) represents the most common commercial utilization of biomass. It is a renewable source of energy and has many advantages from an ecological point of view. Pyrolytic degradation is recognized as an effective method for the production of high yields of char or liquid/gaseous fuels. Pyrolysis is the first stage of both combustion and gasification. So, pyrolysis is not only an independent conversion technology, but also part of the gasification and combustion processes. Pyrolysis consists of a thermal degradation of the initial solid fuel (in the absence of oxygen/air) into: solid (charcoal), liquids (tar and other organics) and gaseous products. Depending on the type of pyrolysis, different species can be obtained. For the purposes of the current study, biomass pyrolysis models may be divided into two primary categories, micro- and macro-particle models. Micro-particle pyrolysis involves the thermal decomposition of virgin matter with sample sizes sufficiently small such that diffusion effects become negligible and the intensity of pyrolysis is controlled with kinetics. Thus, micro-particles are desirable in experiments focusing on identification of kinetic schemes. Critical particle size estimates for kinetic control are generally ~0.1~1 mm and decrease with increasing pyrolysis temperatures [1]. Particles larger than the critical limit are characterized by relatively large diffusion effects which can strongly affect the pyrolysis evolution due to

internal and external temperature gradients, thermal inertia due to heat capacity effects and also temperature variations resulting from endotermic (or exotermic) reactions. Biomass particles commonly have more irregular shapes and much larger sizes than pulverized coal. Larger particle sizes establish the potential for large internal temperature and composition gradients that complicate combustion models. Slab represent a characteristic case because wood chips can be consider as a bunch of small slabs and therefore are especially appropriate for modeling biomass particles. In this article, a microparticle mathematical model of pyrolysis of slab is developed i.e. a kinetic mathematical model. When a solid particle of biomass is heated in an inert atmosphere the following phenomena occur. Heat is first transferred to the particle surface by radiation and/or convection and then to the inside of the particle. The temperature inside the particle increases, causing removal of moisture that is present in the biomass particle, the pre-pyrolysis and main pyrolysis reaction takes place as discussed.. Extensive reviews based on a number of papers done by Di Blasi [3] show that the pyrolysis of biomass involves a complex series of reactions. Most of these observations which are made from reported publications have been mainly focused on woody material and in a few cases on agricultural residues. Pyrolysis in wood is typically initiated at 2000C and lasts till 450-5000C depending on the species of wood. The present study examines how a wood solid generates fuel volatiles in response to a prescribed timevarying heat flux. Kinetic modeling is done to find the thermal behavior of biomass. Biomass from wood processing industries consists of residues and waste from cutting, grinding and planing. Industrial wood residues present significant energetic potential in Serbia. Biomass from wood processing industries can be used as fuel in boilers and for biofuel production. A different mechanism for the pyrolysis of single particle wood and the corresponding kinetic parameters has been reported in literature. In the present work, a two-step kinetic model proposed by Babu et. al. [2] is modified for a slab particle. The proposed model is simulated and results are compared with experimental results. Non-isothermal pyrolysis date were analyzed. This procedure is applied for different values of pyrolysis heating rate. By using the proposed model of this study, it is possible to predict the pyrolysis rate over a wide range of heating rate.

2. Mathematical modeling 2.1. Kinetic model The pyrolysis reaction can be descriebed by means of the following scheme [2]: BIOMASS reaction I

reaction II

( VOLATILE + GASES ) 1 + ( CHAR ) 1

reaction III

( VOLATILE + GASES ) 2 + ( CHAR ) 2

When biomass is heated, it decomposes to volatiles, gases and char. The volatiles and gases may further react with char to produce different types of volatiles, gases and char where the compositions are different. The kinetic equations for the mechanisms shown above: dCB n n = −k1 (CB ) 1 − k2 (CB ) 1 dt

(1)

dCG1 n n n = k1 (CB ) 1 − k3 (CG1 ) 2 (CC1 ) 3 dt

(2)

dCC1 n n n = k2 (CB ) 1 − k3 (CG1 ) 2 (CC1 ) 3 dt

(3)

dCG 2 n n = k3 (CG1 ) 2 (CC1 ) 3 dt

(4)

dCC 2 n n = k3 (CG1 ) 2 (CC1 ) 3 dt

(5)

where,

( (

) )

= k1 A1 exp ( D1 / T ) + L1 / T 2  = k2 A2 exp ( D2 / T ) + L2 / T 2  = k3 A3 exp  − E3 / ( RcT )  .

Adding eqs. (1), (3) and (5) gives, dCB dCC1 dCC 2 dρ + + = −k1CBn1 = dt dt dt dt

(6)

2.2. Heat transfer model Consider a slab of thickness 2L. Assume the length of the two other dimensions of the slab to be so large that heat transfer takes place in the one direction only. d ( −ρ ) d ( c p ρT )  d 2T  k  2  + ( −∆H ) × = dt dt  dx  As ρ and T are function of t, eq.7. can be simplified as: d ( −ρ ) d (T )  d 2T  k  2  + ( −∆H + c pT ) × = ρc p dt dt  dx 

Where:

(7)

(8)

1112 + 4.85 (T − 273 ) cp =

k =+ 0.13 0.0003 (T − 273 ) The thermal properties of the solid (i.e, specific heat, density and thermal conductivity) are presumed to vary continuously from their values for the virgin wood to their values for the char. The initial and boundary conditions for eq.(8) are: Initial conditions: T ( x ,0 ) = T0 t=0 Boundary conditions:  dT  t>0 x = 0,  dx  = 0   x =0  dT  t>0 x= L, −k  = h (Tf − T ) + σε Tf4 − T 4   dx  x = L In last equation, the external heat transfer is considered to occur by a combination of the convective and radiative mechanisms. The governing equations form a set of nonlinear partial differential equations which can be solved numerically. The common explicit forward finite difference method is suitable in this case. This method, however, require the simultaneous solution for the unknowns at each new time step.

(

)

3. Model results Non-isothermal pyrolysis data were analyzed to determine the reaction kinetics for wood particle at two heating rates. Consider a piece of wood, of thickness 1mm, initially at ambient temperature. The piece of wood now exposed to heat flux, that temperature changes from 25 to 6500C, and the average heating rates were: case I-210C/min, case II-320C/min and case III-550C/min. The set of equations are solved numerically subjected to the initial and boundary conditions and the results are presented. Given the symmetry of the problem, only half a particle is simulated. Therefore, while the particle center (x=0) is subjected to conditions of zero gradients on the variables, the external surface (x=L) is exposed to heating. The external heat-transfer coefficient at the particle suface is representative of conditions in reactors. This transfer is made by convective transport and radiation.

Simulations has been made with non-shrinking slab-shaped particles to simulate the effects of reaction temperature. Similar could be done to simulate the effects of particle size, external heat transfer coefficient etc. The main results of the mathematical investigation is the mass loss curve for the wood sample as a function of reaction time and heating rate.

Figure 1a. Mass loss of wood sample as a function of reaction time and heating rate 210C/min

Figure 1b. Mass loss of wood sample as a function of reaction time and heating rate 320C/min

Figure 1c. Mass loss of wood sample as a function of reaction time and heating rate 550C/min

4. Experimental analysis To compare mathematical and experimental results, woody biomass was used as experimental sample. The sample was a mixture of six different kinds of woody biomass. The laboratory facility is designed and constructed in the Institute for Energetics, Process Technique and Protection of Environment at the Faculty of Technical Sciences in Novi Sad [4]. During the experimental investigation of pyrolysis process, initial sample mass was 10g. The mixture samples were placed into the biomass sample container and then into the reactor. The heating process was

performed by using electrical heaters and after achieving the temperature of 6500C, that temperature was maintained within a narrow range around 6500C until the sample mass in the reactor was stabilized. Gaseous pyrolysis products were emitted to the atmosphere through the cooler. After the experiment, liquid phase and char mass were measured.

Figure 2a. Mass loss of wood sample as a function of reaction time and heating rate: 210C/min for model and 240C/min for experiment

Figure 2b. Mass loss of wood sample as a function of reaction time and heating rate: 320C/min for model and experiment

Figure 2c. Mass loss of wood sample as a function of reaction time and heating rate: 550C/min for model and 560C/min for experiment

The process of biomass mixture pyrolysis was carried out with controlling the temperature and mass of the sample. The sample were subjected to the temperature range of 25-6500C, and the average heating rates were 24, 32 and 560C/min. All the experiments were conducted at atmospheric pressure.

5. Conclusions

A mathematical model for single-particle processes has been set and numerically solved to predict the characteristics of the fast pyrolysis of wood particles in reactor. For conditions of interest in fast pyrolysis: particle size of 1mm, a final reactor temperature of 923K and actual heating rates experienced by wood particles are: 21, 32 and 550C/min. In particular the solutions compared the effects of heating rate. Figure 1. shows that for the lower heating rate takes more time to pyrolyse than for higher values of heating rate, for the same size of biomass sample. It is clearly shows that lower heating rate pyrolysis produces more charcoal and higher heating rate of pyrolysis produces more amounts of volatiles and gases and less amount of charcoal. The theory has been compared to experiment. Because wood is a heterogeneous material, the mass loss curves show the existence of several zones associated with the devolatilization of the different components. Following this experimental observation, two-step reaction schemes of primary wood degradation have been proposed to describe the evolution of the reaction zones. Unlike experimental results, mathematical results shows that ones reached temperature about 300-3500C, pyrolysis process start very fast and does not depend much on the temperature change.

Nomenclature CB

concentration of B, kg / m3

k1 , k2 , k3

rate constants, s−1

CC1

concentration of C 1, kg / m3

A1 , A2 , A3

constants, s −1

CC 2

concentration of C 2 , kg / m3

constants, K 2

CG1

concentration of G 1 , kg / m3

L1 , L2

D1 , D2

constants, K

n1 , n2 , n3

orders of reactions

CG 2 cp

concentration of G 2 , kg / m

3

specific heat, J/kg K

h k

convective heat transfer coefficient, W/m K thermal conductivity, W/m K

t ρ0 σ

time, s dencity, kg / m3 Stefan-Boltzmann constant

2

Characteristic values for wood h = 8.4 W / m2

T0

initial temperature, K

T Tf

temperature, K final temperature, K

∆H ε

ρ0 = 650 kg / m3

= A1 9.973 × 10 −5 s −1 D1 = 17254.4 K

n1 = 1 n2 = 1.5

= A2 1.068 × 10 s D2 = 10224.4 K

2L=1 mm T0 =473 K

L1 = −9061227 K 2 −3

−1

L2 = −6123081K 2 = A 5.7 × 105 s −1 3

E3 = 81000J / mol

heat of reaction, kJ/kg emissivity coeficient

n3 = 1.5

Tf =873 K ε =0.95

= σ 5.67 × 10 −8 W / m2 K 4

∆H = −255000J / kg

6. Acknowledgements This research is supported by the Ministry of Education and Science of the Republic of Serbia in the framework of the project “Development of methods, sensors and systems for monitoring of water, air and soil quality” (Project No. III 43008).

7. References

[1] Miller, R.S, Bellan, J: A Generalized Biomass Pyrolysis Model Based on Superimposed Cellulose, Hemicellulose and Lignin Kinetics, Combustion, Science and Technology,Vol.126, 1997, pp.97137. [2] Babu, B.V, Chaurasia, A.S: Modeling for pyrolysis of solid particle: kinetics and heat transfer effects, Energy Conversion and Management, Vol.44, 2003, pp.2251-2275. [3] Di Blasi, C: Modeling chemical and physical processes of wood and biomass pyrolysis, Progress in Energy and Combustion Science, Vol.34, 2008, pp.47-90. [4] Kosanić,T, Ćeranić, M, Đurić, S, Grković, V, Milotić, M, Brankov, S: Experimental Investigation of Pyrolysis Process of Woody Biomass Mixture, Journal of Thermal Science, Vol.23, 2014, pp.290296. [5] Energy Balances of the Republic of Serbia for 2011, Statistical Office of the Republic of Seerbia and Ministry of Mining and Energy, Belgrade,2010. [6] Ministry of Energy and Mining, Renewable Energy Sources in Serbia, http://www.mre.gov.rs/navigacija.php?IDSP=299



CFD Analyses to the Impact of the Pipe Connectors on the Flow Distribution in Parallel Short Pipelines Filip Stojkovskib), Valentino Stojkovskia) (CA), Aleksandar Nospala) a)

University “Ss. Cyril and Methodius” Skopje, Macedonia Faculty of Mechanical Engineering Skopje b) Iskra Impuls l.t.d. Slovenia Branch Office Skopje, Macedonia Abstract: CFD as e technique for developing solutions in complex fluid flow has been used for years, to shorten time and to obtain a global and detailed picture of the flow problem. This analysis is part of a project where the case was to develop a parallel pipe branch connected to the main pipeline to be fulfilled with a certain amount of flow, later to be used as a measuring line of the flow capacity of the system. For this purpose, a few models were built as different variant solutions of the parallel pipe plugs to the main pipeline, so to obtain the flow capacity and the flow distribution in the parallel pipeline system. The parallel pipelines are with different dimension. The numerical models were built in the CAD software Gambit 2.2.30. and the CFD code was solved in Fluent 6.2.16. for various predicted working conditions of the system. The results of the numerical simulations for the different variant solutions are presented in this paper. The analyses of the results can be useful for the designer and can give ideas for further works. Keywords: CFD, Parallel pipelines, Flow distribution, Pipe connectors.

1. Introduction The CFD simulations are made on multiple built models for the purpose of a project for a system of flow measurement of heated water in the heating supply facility for Eastern Skopje. The part of the system is composed from a main measuring pipe line where there is an inflow of water from one side and the other side is with a blind flange at the end of it, 3 outlet pipes (A, B and C) where every pipe has weighted amount of outflowing water and 2 injecting pipes (BP-1 and BP-2) which are injecting water in the distributive collector pipe, fig.1. This system was meant to be upgraded in a way to diverge a certain amount of flow through a new parallel measuring pipe line in case of increased flow capacity in the system.

Figure 1. Injection pipe and distributive collector pipe with flowmeter

The main measuring pipe line has already a flowmeter and it was crucial for the system that the work of the flowmeter would not be disturbed. The parallel measuring line which has to be installed has some initial conditions which have to be taken into concern: - In case of increased flow capacity in the system, the new installed parallel measuring pipe line has to accept those amounts of water flow - The parallel measuring pipe line has to be fitted in the available space of the pump station

-

The parallel measuring pipe line to consider the needs of the new flow meter for exact measurements

With the increase in the capacity of the pump station, the existing distribution pipeline with built-in flowmeter does not allow flow measurement, that is, the flow velocity of the water is above 10 m / s and makes the flow meter useless. Therefore, it is envisaged to extend the measuring line by placing a parallel pipeline to the existing distribution collector, on which another flow meter will be installed, while at the same time ensuring the redistribution of water to both systems. When selecting the technical solution for upgrading the existing system, the limited space conditions for the installation of the system should be taken into account. The models which were built for the purpose of this solution, are technical solutions of the types of pipe connectors (straight branching, elbow suction branch, diffuser ending branch etc.) and they gave different results for the diverged water flow amount in cases of increased water flow capacity in the system.

2. Models of parallel pipeline According to the conditions presented in the introduction of this paper, a few numerical models were built to obtain the suitable solution for the problem. The main goal was to build a parallel pipe system that will take over an amount of flow for the measuring purposes. A numerical model has been established in order to obtain a detailed picture of the flow and distribution of the flows with the collector and parallel pipeline. A numerical CFD calculation approach is chosen for the calculation, because the classical engineering approach (through hydraulic resistance coefficients) can not include the phenomenon of fluid flow phenomena and obstruction of the installed elements of the parallel pipeline. The technical solution for selecting the construction of the parallel pipeline connected to the distribution collector was prepared step by step. The concept of placement of the parallel pipeline contains sections of connection to the distribution pipe (inlet section), direction, right section for setting additional flow meter and connection to the distribution pipeline (outlet section). Five variant solutions for the by-pass pipeline have been developed, namely: Model 1: By-pass parallel pipe with direct ‘T’ connection to the main measuring pipe line, fig.2 Model 2: By-pass parallel pipe with suction pipe elbow inside the main measuring pipe line, fig.3 Model 3: As model 2 upgraded with a diffusor at the exit branch, fig.4 Model 4: As model 3 with extended pipe branches, fig.5 Model 5: As model 3 with a diffusor pipe at the exit branch settled at an angle of 45°, fig.6 The specifics of the concept in the individual analyzed models are the following: - Model 1- inlet and outlet connection a perpendicular to the wall of distributer pipe, this connection is simplest model for upgrading with parallel pipe line - Model 2- inlet section is designed with immersed elbow into the distribution pipeline, positioned in the center of the pipe, but outlet section is the same as Model-1. - Model 3- the concept for inlet section is the same as a Model-2, but outlet section is modified with installing a cone diffuser perpendicular to the wall of distributer pipe - Model 4- this model are the same refer to inlet and outlet connection to the distributer pipe, but the changes are made in the length of cylindrical parts, that mean the distance between axis of the pipe is bigger - Model 5- the concept for inlet section is the same as a Model-2, but outlet section is modified with installing a cone diffuser positioned on 450 to the wall of distributer pipe

Pipe line

Flowmeter M-1

a) Dimension of model -1 b)3D model for numerical calculation Figure 2. Geometrical details of the model 1 Pipe line

Flowmeter M-1


Flowmeter M-1


Flowmeter M-1


Flowmeter M-1

a) Dimension of model -5 b)3D model for numerical calculation Figure 6. Geometrical details of the model 5

3. Boundary condition for numerical calculation The models were built in the CAD software Gambit and the CFD simulations were made in Fluent. The conditions are settled as: a) Numerical model: - Stationary flow - Turbulent model k-ε b) Initial conditions: - Variable input flow in the collector pipe, presented with the average velocity of 2, 3, 4 and 5 m/s. - Variable flow input from the 2 injecting pipes (BP-1 and BP-2) to the collector pipe from 0, 1, 2, 3, 4, 5 and 6 m/s. - Weighted outflow through the 3 pipes (pipe A-42%, pipe B-36% and pipe C-22% - results obtained from the measured exploitation conditions) The estimation of the efficiency of the model variant solutions is made by the amount of flow diverged through the parallel measuring line (DN500) for different operating conditions. The flow input is made from the collector pipe (DN800) by itself and the injection pipes (DN300) (BP-1 and BP-2). The total flow in the measuring system is a sum of these flows. With the upgrade of the parallel measuring line, the flow is divided through the main collector pipe DN800 and the parallel measuring line DN500. Basically, the model solution should provide a flow diversion through the by-pass measuring line DN500 in a limited amount not to overcome the maximal flow needed through the collector pipe DN800 (Qcmax=7600 m3/h) in case of increased flow capacity in the system. From the other hand, in regular operating conditions, the collector pipe and the by-pass measuring line should have flows in the range of the allowance of the flow meters.

4. Results and discussion The results of the numerical calculation are given in the form of diagrams of comparison of exploitation conditions and comparison of variant set solutions for analysis. The main analysis of the results is given in terms of the percentage share of the distribution of the flow through the upgraded parallel pipeline. The conditions of flow distribution for different operating conditions are given as ratio from the total income water flow and the diverged flow through the parallel pipe and as a ratio of the injection pipe flow with the collector flow (Qbp/Qc). 4.1. Model 1 The numerical calculation for the model-1 is exited for two input velocities in the collector pipeline (v c =3 and 4 m/s) and the symmetric flow through the injection pipelines. The percentage involvement of the flow of water in the parallel pipeline is given in Figure 7. 6.00% 5.80% 5.60%

vc=4 m/s

vc=3 m/s

Qv-M2/Qs (%)

5.40% 5.20% 5.00% 4.80% 4.60% 4.40% 4.20%

Qbp1=Qbp2

4.00% 4000

5000

6000

7000

8000

Qs (m3/h)

9000

10000

11000

Figure 7. Flow capacity in the upgraded pipeline in terms of flow in the system in symmetric openness of injection pipelines

On the basis of the obtained calculations, it was concluded that: - in the distribution of the flow, the parallel pipeline participates from 4.9 to 6% of the total flow in the measuring system - such a distribution of the flow between the measuring lines does not satisfy the conditions for measuring the flow in conditions of increasing the capacity of the pump station, because the quantity of water which diverts to the parallel pipe is very small. Model 1 is representing a basic model of parallel pipe with direct connection to the main pipe wall i.e. ‘T’ branching. The results obtained in this model where not satisfactory, but gave the idea for building the model 2, with the suction elbow. 4.2. Comparison of Model 2 to Model 5 Conceptual design models for the parallel pipeline (Model-2 to Model-5) differ in the constructive finishes of setting the solution. The definition of the diverted quantity of water to the parallel pipeline for the various analyzed variant solutions are given in Fig. 8

36% 35% 34%

Qv-M2/Qs (%)

33% 32% 31% 30% T

29%

KD

28%

vc=4 m/s

LD

27%

Qbp1=Qbp2

KD-45

26% 0%

10%

20%

30%

40%

50%

Q-bp/Qc (%)

Figure 8. A comparative diagram for diverting the quantity of water to the parallel pipeline

From the numerical results obtained, we can conclude the following: - the concept of the solution for capturing water to the parallel pipeline by immersing an elbow in the distribution pipe, provides an increased diversion of the amount of water to the parallel pipeline - in the distribution of the flow, the upgrade for a variant of 2 to 5 participates from 29 to 35% of the total flow in the measuring system - such a distribution of the flow between the measuring lines meets the conditions for measuring the flow in conditions of increasing the capacity of the pump station - the most favorable distribution is achieved by the proposed variant solution -5. The diversity in the distributive ability of the water in the solutions considered may exist due to the diversity of the flow conditions in the inlet and outlet cross-section of the parallel pipeline to the distribution pipeline. The stream flow images which are provided on the input and output cross-section of the connection are given in Figure 9.

MODEL-2-4-6-6

MODEL-3-4-6-6

MODEL-4-4-6-6

KODEL-5-4-6-6 Figure 9. Stream flow field at the inlet and outlet cross-section

From the numerical calculated flow field image, presented through the field of velocity, the following can be concluded: - The velocity of the fluid flow field (for the same conditions of exploitation) at the inlet section of the parallel pipeline has an insignificant change (differences) - The velocity of the fluid flow field (for the same conditions of exploitation) at the exit section of the parallel pipeline has a significant change in the outcome conditions - The difference in the efficiency of the quantity of diverted water to the parallel pipeline is mainly the result of the conceptual solution of the water out flow to the distribution pipe.

4.3. Capability of Model 5 The most favorable technical conditions for the diversion of water to the parallel pipeline are obtained for the construction given in Model 5. According to the current conditions of exploitation, the flow of water to the distribution pipeline can be different, i.e. it is possible to have a flow only from the main tube or from the main tube and injector pipes. On fig. 10 and fig.11 are gives the numerical results for the diverted quantity of water in conditions the velocity at main pipe v c =2, 3, 4 and 5 m / s and with simultaneous symmetrical water flow through the injectors pipes.

37%

36%

vc=3 m/s vc=4 m/s

35%

Qv-M2/Qs (%)

34%

Qv-M2/Qs (%)

vc=2 m/s

36%

35%

33% 32% vc=2 m/s 31%

vc=3 m/s

vc=5 m/s 34% 33% 32% 31%

vc=4 m/s

Qbp1=Qbp2

30%

Qbp1=Qbp2

30%

vc=5 m/s

29%

29% 0

2000

4000

6000

8000

10000

12000

0%

14000

40%

20%

Figure 10. Distribution of water in a parallel pipeline vs discharge in the system

60%

80%

100%

Qbp/Qc (%)

Qs (m3/h)

Figure 11. Distribution of water in a parallel pipeline vs relative discharge between injected and main pipe flow

From the obtained results we can conclude: • The diverted amount of water to the parallel pipeline has variable intensity depending on the conditions of the system's exploitation • Regardless of the velocity of water in the main pipeline, the diverted amount of water to the parallel pipeline is in the range of 31 to 35% • The diverted amount of water to the parallel pipeline represented in relation of the relative flow between the injected and the flow in the main pipe shows continuous dependence on the operation of the system. The influence of the intake of flow through the first injection pipe BP-1 or the second injection pipe BP-2 on the effect of the diverted water quantity to the parallel pipeline is given on Figure 12. 37% Qbp1(1); Qbp2(0)

36%

Qbp1=Qbp2 Qbp1(0); Qbp2(1)

36%

Qv-M2/Qs (%)

35% 35% 34% 34% 33%

vc=4 m/s

33% 32% 0%

10%

20%

30%

40%

50%

60%

Q-bp/Qc (%)

Figure 12. Influence of active injection pipe on water distribution in a parallel pipeline

From the obtained results we can conclude: - the exploitation of injector pipelines affects the distribution of the flow between the measurement lines of the upgraded measuring system - the more frequent effect of the flow distribution is obtained under water flow conditions through the first injection pipeline (BP-1). - seen in a wider range of exploitative use of the system, more favorable conditions (more stable) are achieved through the symmetrical use of injection pipelines. The difference in the diverted amount of water under different operating conditions of the injection tubes (BP-1 and BP-2) is the result of the different fluid flow conditions at the inlet section of the system. Figure 13 shows the images of velocity fluid flow field obtained during the different activity of the injection pipes. There are presented the condition with inflow in the main pipe (v c (1)), inflow through the injection pipe one (Qbp-1(1)), inflow through the injection pipe two (Qbp-2(1)) and variation of the exploitation conditions. 8.00e+00

8.00e+00

7.60e+00 7.20e+00 6.80e+00 6.40e+00 6.00e+00 5.60e+00 5.20e+00 4.80e+00

7.60e+00 7.20e+00 6.80e+00 6.40e+00 6.00e+00 5.60e+00 5.20e+00 4.80e+00 4.40e+00 4.00e+00 3.60e+00 3.20e+00 2.80e+00 2.40e+00 2.00e+00 1.60e+00 1.20e+00 8.00e-01 4.00e-01 0.00e+00

4.40e+00 4.00e+00 3.60e+00 3.20e+00 2.80e+00 2.40e+00 2.00e+00 1.60e+00 1.20e+00 8.00e-01 4.00e-01

Y

0.00e+00

X Z

Y

X Z

v c (1), Qbp-1(1); Qbp-2(1); 8.00e+00 7.60e+00 7.20e+00 6.80e+00 6.40e+00 6.00e+00 5.60e+00 5.20e+00 4.80e+00 4.40e+00 4.00e+00 3.60e+00 3.20e+00 2.80e+00 2.40e+00 2.00e+00 1.60e+00 1.20e+00 8.00e-01 4.00e-01 0.00e+00

Y

v c (1), Qbp-1(0); Qbp-2(0); 8.00e+00 7.60e+00 7.20e+00 6.80e+00 6.40e+00 6.00e+00 5.60e+00 5.20e+00 4.80e+00 4.40e+00 4.00e+00 3.60e+00 3.20e+00 2.80e+00 2.40e+00 2.00e+00 1.60e+00 1.20e+00 8.00e-01 4.00e-01 0.00e+00

X Z

v c (1), Qbp-1(0); Qbp-2(1);

Y

X Z

v c (1), Qbp-1(1); Qbp-2(0);

Figure 13. Influence of active injection pipe on water distribution in a parallel pipeline

From the numerical calculated flow field image, presented through the field of velocity, the following can be concluded: - The different conditions of water flow to the system creates also various fluid flow fields on the inlet section of the parallel pipeline - The numerical results show that a more favorable fluid flow field which resulting in a higher percentage of water diversion to the parallel pipeline is obtained in the case of the injection from pipe one (Qbp-1 (1)), as the injection pipe two is closed (Qbp-2 (0))

Conclusion According to the obtained results, it can be concluded that Model 5 gave the best results for the problem. The model 1 was the first idea how to develop the system further to end with model 5. The idea was developed to obtain stable flow distribution in the needed amount. The results of the numerical calculation given in this paper show that small differences in the conceptual design at short pipelines have a significant influence on the technical hydraulic characteristics of the projected system. The ideal solution for increasing the flow through the parallel pipeline, by inserting internal elbow, applied in the project solution, provides a sufficient balance of the flow in the measuring system. The way of connecting and installing a short parallel pipeline,

where the flow of water plays a major role in defining hydraulic resistance, has a significant impact on defining the technical hydraulic characteristics to be achieved. Technical solutions where the flow phenomena have a dominant influence on the choice of the conceptual solution, and at the same time to achieving the required technical conditions (parameters), can be obtained only through the use of the CFD approach for analysis. The results and considered variant solutions can be used as the basis for future design solutions that would be used when designing such or similar systems.

Symbols Qs

–

System water flow

Qv-M1

–

Volumetric water flow through the main measuring pipe line DN800

Qv-M2 Qc

– –

Volumetric water flow through the new parallel measuring pipe line DN500 Volumetric water flow into the main pipeline DN800

Qbp-1

–

Volumetric water flow injected through BP-1

Qbp-2

–

Volumetric water flow injected through BP-2

Qbp-1 = Qbp-2

–

vc v bp

– –

Average velocity through the main measuring pipe line Average velocity of the pipe injector BP-1 and BP-2

–

Flow ratio of the volumetric flowrate through the by-pass measuring line and the system flowrate

–

Flow ratio of the injected flow through the injector pipes and the main measuring pipe line

Qv-M2/Qs Qbp/Qc

Symmetrical flow injected through BP-1 and BP-2

References [1] [2] [3] [4]

V.Stojkovski, Z.Kostikj, F.Stojkovski,: Upgrading the measurement pipelines in the pump station TO Istok, Faculty of Mechanical Engineering, Skopje, Report No.ST-02/17, 2017 T.Bundalevski,: Fluid mechanic, (book on Macedonian language), MB-3, 1995 A.Nospal: Fluid flow measurement and instrumentation, (book on Macedonian language), MB-3, 1995 F.Corlukic: Merenje protoka fluida, (book on Croatian language), Tehnicka kniga, Zagreb, 1975



Numerical Optimization of Pulverized Coal Furnace Sorbent Injection Under Various Operating Conditions Ivan Tomanovića, Srđan Beloševićb, Aleksandar Milićevićc, Nenad Crnomarkovićd and Dragan Tucakoviće a

University of Belgrade, „Vinča“ Institute of Nuclear Sciences, Belgrade, RS, [email protected] b c

University of Belgrade, „Vinča“ Institute of Nuclear Sciences, Belgrade, RS, [email protected]

University of Belgrade, „Vinča“ Institute of Nuclear Sciences, Belgrade, RS, [email protected] d

e

University of Belgrade, „Vinča“ Institute of Nuclear Sciences, Belgrade, RS, [email protected]

University of Belgrade, Faculty of Mechanical Engineering, Belgrade, RS, [email protected]

Abstract: Calcium based sorbent injection in pulverized coal utility boiler furnace for SO 2 emission reduction process optimization study is performed by means of an in-house developed software tool. An Euler-Lagrangian approach to fluid-particles motion in two-phase turbulent reactive flow is used, and phases are coupled by PSI Cell method, with radiative heat exchange taken into account by six-flux radiation model, while the combustion reaction model is based on available experimental data on domestic lignites. Sorbent particles and their reactions of calcination, sintering and sulfation are modeled by grain model, taking into account internal particle structure. Models of processes in the boiler furnace have been previously verified, and simulation results are used for analysis and determining of optimal directions in which the process should be controlled in order to achieve optimal emission reduction, and at the same time to preserve internal sorbent particle structure allowing for its further use in combined desulfurization technologies. Attention is given to operating conditions under various fuels combustion and changes in flame geometry, and the influences they exert on SO 2 content at the furnace exit are observed as well. Keywords: Numerical simulation, Utility boiler, Ca-based sorbent, Desulfurization Corresponding author: Ivan Tomanović, [email protected]

1. INTRODUCTION It is well known that uncontrolled sulfur emission from power plant boilers [1] is a great ecological problem and must be addressed properly by introducing technologies and processes [2]–[12] in order to control and limit the amount of sulfur released in atmosphere. Legal documents considering emissions and limitations exist on both global and local levels. In European Union, such document is “Directive 2010/75/EU of the European Parliament and of the Council of 24 November 2010 on industrial emissions (integrated pollution prevention and control)” which repealed the “Directive 2001/80/EC of the European Parliament and of the Council of 23 October 2001 on the limitation of emissions of certain pollutants into the air from large combustion plants”, while in Republic of Serbia emission limit values are set in “Uredba o uslovima za monitoring i zahtevima za kvalitet vazduha” („Сл. гласник РС“, бр 11/10, 75/10, 63/13).

Desulfurization technologies in use today vary in complexity, efficiency and cost. Selecting optimal technology for every plant depend highly on regulations that must be met, expected life of plant, installation space for new equipment, availability of materials. Most common are wet scrubbing technologies [9]–[11], which have high efficiency. Alternative to them are sorbent injection processes, with lower efficiencies [2], [7], [8], [12], but are easier to install on existing plants and have lower initial costs. Also, sorbent injection can be used together with scrubbing technologies, decreasing water consumption and in certain cases improving plant efficiency due to fact that sorbent undergoes calcination process in furnace itself. In this paper we approach the problems of optimization of direct in-furnace calcium based sorbent injection particles by means of mathematical modeling. Previously developed model [13]–[15] and software [16]–[18] are used in simulations. Due to complexity of processes in boiler furnace it is hard to predict sorbent injection behavior in such system, and lack of process optimization can lead to great variations in process efficiency [19]. Numerical analysis allow us to perform in-depth analysis of these processes, and estimate the effects the variations in process configuration and parameters would have at furnace exit, considering content of SO 2 and other relevant parameters under wide range of operating conditions. In order to predict proper process configuration for clean and efficient boiler operation, working conditions under various fuels combustion and changes in flame geometry, and the influences they exert on SO 2 content at the furnace exit are closely observed.

2. MATHEMATICAL MODEL Well known Euler-Lagrangian approach to two phase modeling is used. Model was previously well described in papers [20], [21], and here are highlighted the most important properties of the model. Model uses Euler-Lagrangian approach to two-phase gas-particle flow modeling. Conservation equations for mass, momentum, energy, mixture components concentration, turbulent kinetic energy and dissipation are described with time-averaged Eulerian equations, while the disperse phase and changes in particle motion, energy and mass are described by corresponding equations in Lagrangian field. Closure of time averaged Navier-Stokes equations is done by k-ε turbulence model. Coal particle interaction with gas is described by PSI Cell method, while the gas influence on particles is considered by diffusion model of particles turbulent dispersion. Six-flux model is used for radiative heat transfer modeling. Sorbent particle reactions with gas also use PSI Cell approach, while the changes in particle structure and resulting reaction rates are modeled using PSS Model [22]–[25]. Sorbent particles are considered to react in two steps. The first one is calcination of particle (during which the CaCO 3

contained in particle transforms into CaO), during which the particle develops internal surface. This goes together with sintering which is opposed to previous process and leads to loss of internal surface. Second step is sulfation reaction, during which, particle captures SO 2 from surrounding gas.

3. NUMERICAL SIMULATIONS RESULTS Numerical simulations are performed on pulverized lignite powered boiler furnace Kostolac B2 350 MW e . This furnace was also used in previous research, and various modifications and their influence were examined on it [20], [21]. Furnace itself has eight burners tangentially placed around the furnace center, two at each wall. Each burner has four tiers – two main and two upper-stage burners.

Burner in use Yes Yes Flow through burner Primary air 43,726 43,726 [kg/s] Coal dust 10,385 10,385 [kg/s] Secondary air 31,793 31,793 [kg/s] Core air 6,432 6,432 [kg/s] Air mixture Oxygen 9,220 9,220 [%] Nitrogen 58,095 58,095 [%] Carbon-dioxide 9,772 9,772 [%] Water vapor 22,758 22,758 [%] Secondary air distribution over burner tiers UP-U 10,366 10,366 [%] UP-M 6,121 6,121 [%] UP-L 10,648 10,648 [%] MB-U 16,642 16,642 [%] MB-UL 16,848 16,848 [%] MB-LU 17,048 17,048 [%] MB-L 17,294 17,294 [%] Core air distribution over burner tiers UP-U 24,570 24,570 [%] UP-L 24,570 24,570 [%] MB-U 25,115 25,115 [%] MB-L 25,745 25,745 [%] Air mixture distribution over burner tiers UP-U 22,014 22,014 [%] UP-L 20,930 20,930 [%] MB-U 26,795 26,795 [%] MB-L 30,261 30,261 [%] Coal dust distribution over burner tiers UP-U 10,500 10,500 [%] UP-L 19,500 19,500 [%] MB-U 24,500 24,500 [%] MB-L 45,500 45,500 [%]

Burner 6

Burner 5

Burner 4

Burner 3

Burner 2

Burner 1

Burner 8

Burner 7

Unit

Table 1. Input data for furnace simulation for test-cases based on coals 6489 and 7327

Yes

Yes

Yes

Yes

Yes

No

43,726 10,385 31,793 6,432

43,726 10,385 31,793 6,432

43,726 10,385 31,793 6,432

43,726 10,385 31,793 6,432

43,726 10,385 31,793 6,432

0,000 0,000 8,356 1,439

9,220 58,095 9,772 22,758

9,220 58,095 9,772 22,758

9,220 58,095 9,772 22,758

9,220 58,095 9,772 22,758

9,220 58,095 9,772 22,758

0,000 0,000 0,000 0,000

10,366 6,121 10,648 16,642 16,848 17,048 17,294

10,366 6,121 10,648 16,642 16,848 17,048 17,294

10,366 6,121 10,648 16,642 16,848 17,048 17,294

10,366 6,121 10,648 16,642 16,848 17,048 17,294

10,366 6,121 10,648 16,642 16,848 17,048 17,294

8,876 5,247 9,052 14,201 14,348 14,485 14,642

24,570 24,570 25,115 25,745

24,570 24,570 25,115 25,745

24,570 24,570 25,115 25,745

24,570 24,570 25,115 25,745

24,570 24,570 25,115 25,745

24,715 24,715 25,000 25,569

22,014 20,930 26,795 30,261

22,014 20,930 26,795 30,261

22,014 20,930 26,795 30,261

22,014 20,930 26,795 30,261

22,014 20,930 26,795 30,261

0,000 0,000 0,000 0,000

10,500 19,500 24,500 45,500

10,500 19,500 24,500 45,500

10,500 19,500 24,500 45,500

10,500 19,500 24,500 45,500

10,500 19,500 24,500 45,500

0,000 0,000 0,000 0,000

Secondary air flow: UP-U – above upper-stage upper burner; UP-M – between upper-stage burners; UP-L – below upper-stage lower burner; MB-U – above upper main burner; MB-UL – below upper main burner; MB-LU – above lower main burner; MB-L – below lower main burner;

Coal, air mixture and core air flows: UP-U – upper-stage upper burner; UP-L – upper-stage lower burner; MB-U – upper main burner; MB-L – lower main burner.

The furnace model, in these simulations, is modified in such manner to allow for direct sorbent injection in boiler furnace. Pulverized sorbent is injected through fuel feeding line together with pulverized lignite. Amount of sorbent through each burner tier is varied, depending on amount of combustible sulfur in fuel in order to keep target calcium-sulfur ratio. Coal and air flows in simulation test-cases are given in table 1, while the proximate and ultimate analysis of coal, as received, can be found in table 2. As we can see, all test-cases have similar coal and air flows, but the differences can be found in coal composition, leading to variations in temperatures and pollutant content at furnace exit, and overall different flame geometry inside of furnace. Table 2. Input data for furnace simulation for test-cases based on coals 6489 and 7327 Fuel Proximate analysis Ultimate analysis A coal w coal LHV coal C coal H coal [%] [%] [kJ/kg] [%] [%] 6489 40,7687 8,8000 6489,5 31,9534 3,3730 7327 36,5196 8,8310 7326,9 36,5196 3,4471

O coal [%] 12,4653 12,5201

S comb , coal [%] 1,1406 1,0406

Results of numerical simulations for selected test-cases are given in table 3. First two test-cases, 1. and 2, are base test-cases without sorbent injection. They are used as reference when comparing values at furnace exit, process influence on gas temperature, SO 2 and NO x content reduction. Table 3. Furnace exit gas temperature, SO 2 content, and NO x content

Test-case no.

1. 2.

6489-0 7327-0

3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

6489-8 6489-9 6489-10 6489-11 6489-12 6489-13 7327-8 7327-9 7327-10 7327-11 7327-17 7327-22 7327-23 7327-24

*

SO 2 Sorbent distribution over burner tiers Gas temperature content Upper-stage Main burners burners lower upper lower upper at furnace exit [%] [%] [%] [%] [°C] [mg/Nm3] Base test-cases No sorbent injection 937,34 8127,22 No sorbent injection 1028,76 6815,63 Sorbent injection test-cases* 0 0 0 100 927,57 6688,63 0 0 100 0 943,50 5961,33 0 100 0 0 946,29 5251,44 100 0 0 0 944,02 4563,94 0 33 33 33 939,26 5285,34 25 25 25 25 937,50 3834,63 0 0 0 100 1022,02 5566,54 0 0 100 0 1039,60 4657,53 0 100 0 0 1041,00 4049,02 100 0 0 0 1033,90 3758,52 0 33 33 33 1037,33 3865,79 25 25 25 25 1029,70 2713,88 35 35 15 15 1032,70 2544,43 40 30 20 10 1035,90 2596,67

NO x content

Test-cases with sorbent injection have calcium/sulfur molar ratio Ca/S=2, with sorbent distribute among burner tiers as noted in table.

[mg/Nm3] 327,14 427,86 329,47 354,82 350,16 279,26 344,72 318,02 428,43 449,62 438,47 364,95 438,46 430,51 425,15 424,68

Below these are test-cases with sorbent injection through individual burner tiers, 3-6, 9-12, through three, 7. and 13, and four tiers, 8. and 14. Final two test-cases, 15. and 16, are used to estimate effects of uneven distribution of sorbent over burner tiers, with other parameters unchanged. Upon a closer look at SO 2 content at furnace exit in test-cases with sorbent injection through individual burner tiers, it can be seen that sorbent injection through main burners yields better reduction compared to injection through upper-stage burners. As the sorbent injection position moves from upper stage burners, 3, 4, 9, and 10, down to main burners, 5, 6, 11, and 12, the amount of removed SO 2 at furnace exit increases. Trend in SO 2 content decrease is shown on figure 1. This can be attributed to better targeting of SO 2 rich zones with sorbent particles, as those zones tend to appear closer to flame center (as can be seen from figure 2 for coal 6489 and calcium-sulfur ratio 2), higher temperatures which lead to faster sorbent particle internal surface development and faster sulfation reaction up to a certain point when the sintering takes over. Between test-cases 7327-10 and 7327-11, 11. and 12, there is noticeable drop in process efficiency. Due to high temperatures the sorbent is exposed to in test-case 7327-11, when injected mostly in flame core, rapid sintering occurs, leading to significant decrease in sulfation reaction rate, and renders sorbent particles useless for further application.

9E+03

Fuel 6489

8E+03

7327

𝐒𝐎𝟐 content

7E+03 6E+03 5E+03 4E+03 3E+03

1

2

3

4

5

Sorbent injection vertical position 1 – no sorbent injection; sorbent injected through: 2 – upper-stage upper burners, 3 – upper-stage lower burners, 4 – main upper burners, 5 – main lower burners

Figure 1. Predicted sulfur dioxide content at furnace exit based on calcium - sulfur molar ratio of 2,00 for test-cases based on fuels 6489 and 7327 (table 3 1.-7., 9.-12.)

From these test-cases it is noticed that sorbent injection through lower burner tiers yields better SO 2 capture. However, in all test-cases the sorbent particle sintering, which highly depend on local gas temperature, must be taken into account when determining best injection position. This is especially true if the sorbent is to be used for desulfurization after it leaves the boiler furnace. Thus some testcases, even though they have best SO 2 content reduction, should be considered with caution, as the sorbent particles could lose internal reactive surface rapidly, rendering them useless for further processes. When considering optimal test-cases good balance between SO 2 content reduction and particle sintering must be obtained, allowing for further sorbent use, and capturing significant amount of SO 2 at the same time.

a

b

c

d

Injection position: a – upper-stage upper burner, b – upper-stage lower burner, c – main upper burner, d – main lower burner

Figure 2. Influence of sorbent injection on SO 2 concentration in furnace (test-cases table 3 3-6)

Figure 3. Sorbent injection through all burner tiers and temperature and SO 2 concentration (test-case table 3 8)

In order to partially overcome particle sintering issue sorbent can be redistributed among burner tiers, leading to less particles being exposed to high temperatures, and better mixing of sorbent

particles with SO 2 in zones rich with sulfur oxide. This is tested in test-cases 7, 8 (figure 3), 13-16, as previously mentioned. In test-cases ord.no 13-16. it can be noticed that increase in amount of sorbent injected through main burners lead towards better emission reduction, but if excessive amount of sorbent is redirected through lower main burners, sintering influences overall result and leads to slight decrease of process efficiency, with other negative effects that occur in test-case 16. compared to test-case 15. This trend can be seen in figure 4, further stressing the importance of analysis and fine optimization of process under various operating conditions. 2.75E+03

Fuel 7327

𝐒𝐎𝟐 content [mg/Nm3]

2.70E+03

2.65E+03

2.60E+03

2.55E+03

2.50E+03

25

30

35

40

Sorbent % through lower main burners

Figure 4. Predicted sulfur dioxide content at furnace exit based on calcium - sulfur molar ratio of 2,00 and varied sorbent distribution over burner tiers for test-cases based on fuel 7327 (test-cases table 3 14-16)

Finding an optimal setup for sorbent injection is multidimensional problem. Main accent is on sorbent distribution and proper mixing with SO 2 released by combustion of coal. When good mixing is achieved, the increase in process efficiency is noticeable (for different injection positions sorbent penetration in SO 2 rich zones can be seen on figure 2). Second influential parameter is local temperature in zones covered by sorbent particles. The temperature influences not only the sulfation reaction rate, but also calcination and sintering. Extremely high temperatures can lead to decrease in process efficiency, as the sintering becomes dominant process and the loss of particle reactive surface becomes significant. This effect was observed in test-case 16 where it caused increase in emission and particle deactivation due to extensive sintering. The amount of sorbent injected depends on Ca/S molar ratio, and is proportional to combustible sulfur content in coal. Process optimization should always be tried by changes in injection configuration, and the Ca/S molar ratio increase should be considered as last resort, as it directly influences the amount of sorbent spent.

Particle utilization is also an important factor when considering process as whole, and it is directly influenced by other previously mentioned conditions in boiler furnace.

4. CONCLUSIONS Direct furnace injection is interesting technology that can be used in conjunction with scrubbing technologies, or as individual technology on smaller plants that satisfy certain conditions according to legal regulations, such as having working hours during the year under given limit. In order to describe calcination, sintering and sulfation reactions, a model was selected, adapted and incorporated into the in-house developed comprehensive combustion code and used to investigate complex process of desulphurization by direct sorbent injection in the case-study boiler furnace. It has been shown that furnace sorbent injection strongly depends on local conditions inside of furnace, and that great improvement in process efficiency can be achieved by proper distribution of sorbent, depending on local conditions. Injection of sorbent through more than one burner tier leads to improvement in process efficiency; also, targeting zones with high sulfur content improves sorbent utilization and leads to higher process efficiencies. Uneven distribution among burner tiers helps in local calcium-sulfur rate increase in sulfur rich zones. The importance of process planning for various fuels, in this case coals with relatively low LHV (6489) and somewhat higher LHV (7327), is further stresses. It is shown that, for the fuel 7327, test-case 16. has somewhat lower efficiency compared to test-case 15, while the two corresponding test-cases for (6489) do not have that problem, as the sintering does not overcome positive effects of high temperatures, due to lower flame core temperature and slight differences in flame geometry. Changes made to accommodate sorbent injection technology in simulations show no significant influence on changes in nitrogen emission.

ACKNOWLEDGEMENTS This work has been supported by the Republic of Serbia Ministry of Education, Science and Technological Development (project: “Increase in energy and ecology efficiency of processes in pulverized coal-fired furnace and optimization of utility steam boiler air preheater by using in-house developed software tools”, No. TR-33018).

REFERENCES [1] [2]

[3]

Đuković J., Bojanić, V., Aerozagađenje : pojam, stanje, izvori, kontrola i tehnološka rješenja, Mrlješ, Beograd, 2000 Cheng, J.,Zhou, J., Liu, J., Zhou, Z., Huang, Z., Cao, X., Zhao, X., Cen, K., Sulfur removal at high temperature during coal combustion in furnaces: a review, Progress in Energy and Combustion Science, 29 (2003), 5, pp. 381-405 Ostrop, F.-G., Coal Quality Control for the Consumers Along the North-South Railway Line, Bulk Solids Handling, 14 (1994), pp. 87-87

[4]

[5] [6] [7]

[8] [9] [10] [11] [12]

[13] [14]

[15]

[16]

[17]

[18]

[19] [20] [21]

[22]

[23] [24] [25]

Cheng, J., Liu, J., Wu, X., Tang, L., Li, N., Yao, Q., Cao, X., Fang, M., Luo, Z., Cen, K., The characteristics of fluorine and sulfur release during coal combustion, Proceedings, International Joint Power Generation Conference, Burlingame, CA, USA, July, 25-28, 1999 Kargi, F., Robinson, J. M., Removal of Sulfur Compounds from Coal by the Thermophilic Organism Sulfolobus acidocaldarius, Applied and Environmental Microbiology, 44 (1982), 4, pp. 878-883 Kargi, F., Robinson, J. M., Biological removal of pyritic sulfur from coal by the thermophilic organismSulfolobus acidocaldarius, Biotechnology and Bioengineering, 27 (1985), 1, pp. 41-49 Sage, P. W., Ford, N. W. J., Review of sorbent injection processes for low‐cost sulphur dioxide control, Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy, 210 (1996), 3, pp. 183-190 Steward, F. R., Couturier, M., Morris, K., Firing coal and limestone in a low-NO x burner, to reduce SO 2 emissions., Journal of the Institute of Energy, 65 (1992), 465, pp. 177-184 Olausson, S., Wallin, M., Bjerle, I., A model for the absorption of sulphur dioxide into a limestone slurry, The Chemical Engineering Journal, 51 (1993), 2, pp. 99-108 Jozewicz, W., Rochelle, G. T., Fly ash recycle in dry scrubbing, Environmental Progress & Sustainable Energy, 5 (1986), 4, pp. 219-224 Rice, R. W., Bond, G. A., Flue gas desulfurization by in-duct dry scrubbing using calcium hydroxide, AIChE Journal, 36 (1990), 3, pp. 473-477 Zhou, W., Maly, P., Brooks, J., Nareddy, S., Swanson, L., Moyeda, D., Design and Test Furnace Sorbent Injection for SO 2 Removal in a Tangentially Fired Boiler, Environmental Engineering Science, 27 (2010), 4, pp. 337-345 Tomanović, I., Belošević, S., Milićević, A., Tucaković, D., Modeling of calcium-based sorbent reactions with sulfur dioxide, Journal of the Serbian Chemical Society, 80 (2015), 4, pp. 549-562 Tomanović, I., Belošević, S., Stojanović, A., Tucaković, D., Živanović, T., Alternative modeling approaches to sulfation reactions of calcium based sorbents injected in the pulverized coal furnace, Proceedings, 16th Symposium on Thermal Science and Engineering of Serbia,Sokobanja, Serbia, October 22-25, 2013, pp. 65-72 Tomanović, I., Milićević, A., Stojanović, A., Belošević, S., Tucaković, D., Ca-based sorbent reactions modelling for SO 2 emission reduction in pulverized coal combustion conditions, Proceedings, International Conference “Power Plants 2014,” Zlatibor, Serbia, October 28-31, 2014, pp. 704-711 Tomanović, I., Belošević, S., Milićević, A., Tucaković, D., Influence of sorbent particle dispersion in pulverised coal-fired furnace on desulphurization process efficiency, Proceedings, 17th Symposium on Thermal Science and Engineering of Serbia, SIMTERM 2015, Sokobanja, Serbia 20-23, 2015, pp. 228-235 Tomanović, I., Belošević, S., Milićević, A., Crnomarković, N., Tucaković, D., Furnace sorbent injection for SO 2 removal and the effects on combustion and NOx emission - a numerical study, Proceedings, International Conference Power Plants 2016, Zlatibor, Serbia, November 23-26, 2016, pp. E2016-61 Belošević, S., Tomanović, I., Beljanski, V., Tucaković, D., Živanović, T., Numerical prediction of processes for clean and efficient combustion of pulverized coal in power plants, Applied Thermal Engineering, 74 (2015), pp. 102–110 Case, P., Heap, M., Lee, J., McKinnon, C., Payne, R., Pershing, D. W., Use of sorbents to reduce SO 2 emissions from pulverized-coal flames under low-NO x conditions (Progress report),” Cincinnati, Ohio, 1983 Belosevic, S., Sijercic, M., Crnomarkovic, N., Stankovic, B., Tucakovic, D., Numerical Prediction of Pulverized Coal Flame in Utility Boiler Furnaces, Energy & Fuels, 23 (2009), 11, pp. 5401-5412 Belosevic, S., Beljanski, V., Tomanovic, I., Crnomarkovic, N., Tucakovic, D., Zivanovic, T., Numerical Analysis of NO x Control by Combustion Modifications in Pulverized Coal Utility Boiler, Energy & Fuels, 26 (2012), 1, pp. 425-442 Alvfors, P., Svedberg, G., Modelling of the sulphation of calcined limestone and dolomite—a gas-solid reaction with structural changes in the presence of inert solids, Chemical Engineering Science, 43 (1988), 5, pp. 11831193 Alvfors, P., Svedberg, G., Modelling of the simultaneous calcination, sintering and sulphation of limestone and dolomite, Chemical Engineering Science, 47 (1992), 8, pp. 1903-1912 Lindner, B., Simonsson, D., Comparison of structural models for gas-solid reactions in porous solids undergoing structural changes, Chemical Engineering Science, 36 (1981), 9, pp. 1519-1527 Kocaefe, D., Karman, D., Steward, F. R., Interpretation of the sulfation rate of CaO, MgO, and ZnO with SO 2 and SO 3 , AIChE Journal, 33 (1987), 11, pp. 1835-1843



Numerical Simulation of Water Hammer in Penstock of the Hydropower Plant Bistrica Jovana Spasića (CA), Živojin Stamenkovićb, Dragica Milenkovićc a

Faculty of Mechanical Engineering, Nis, Serbia, [email protected] Faculty of Mechanical Engineering, Nis, Serbia, [email protected] c Faculty of Mechanical Engineering, Nis, Serbia, [email protected] b

Abstract: Water hammer is known as the phenomenon of large changes in pressure in the hydraulic systems, resulting from the sudden, controlled and uncontrolled change of the discharge through the system. This phenomenon can affect the safety and reliability of the plant as well as its operating life. Numerical simulations of water hammer can be used in a quick and reliable way to analyze and check the efficiency of the solution so as to protect plants from water hammer in different operating regimes. In this paper numerical simulations of water hammer were conducted and the results obtained for the various operating regimes of the Hydropower Plant Bistrica were analyzed. The Hydropower Plant Bistrica has two Francis turbines that operate on a gross head of 376 m with a total installed power of 104 MW. Water is supplied to the turbines over the tunnel with the diameter of 4 m and two parallel penstocks with the length of 8000 m and the diameter which varies from 2.2 m to 2.0 m. A surge tank is located at the beginning of the penstock and used for the protection against water hammer. Numerical simulations of water hammer were conducted in the AFT Impulse program and obtained results are presented graphically with appropriate conclusions. Keywords: Water Hammer, Hydropower Plant, Penstock, Numerical simulation

1. Introduction In case of sudden and uncontrolling changes of the working fluid flow velocity through Hydropower Plant (HPP) changes in pressure in the system can occur which can affect the safety and reliability of the entire HPP, as well as its lifetime. Knowing the transient and the non-stationary processes, as well as the conditions that lead to them, is a good basis for designing systems for controlling plants, devices and security elements for their protection. Many researchers have attempted to obtain the simulation results of transient flow in penstock systems with different methods. Among the approaches proposed to solve the single-phase water hammer equations are Finite Volume (FV), Finite Differences (FD), Finite Elements (FE), Wave Characteristic Method (WCM), and the Method of Characteristics (MOC) [1]. Numerical simulations represent a good way to study non-stationary flow regimes along with the corresponding experimental data. In doing so, it tends to make models as simpler and more comprehensive as possible. The results of numerical analyzes can serve as a basis for selecting reliable operating regimes of the plant, safety devices and protection elements. Numerical simulations can be used to check the operating conditions of existing plants and assess the choice of safety equipment. Knowing the dynamics of transient processes is of particular importance for preventing or reducing the impact of water hammer on the operating parameters of the system and functionality of the plant. Izquierdo and Iglesias developed a computer program to simulate hydraulic transients in a simple pipeline system [2]. They presented a powerful tool for planning the potential risks to which an installation may be exposed and for developing suitable protection strategies. Their model produced good numerical results within the accuracy of the used data. Filion and Karney [3] proposed a method that combined a numerical integration method with a transient simulation model to improve the accuracy and capabilities of extendedperiod simulations in pipe networks. Water distribution for different closing time of main turbine inlet valve using a transient model was analyzed by this method, and then a modified Euler’s method was used to predict the behavior of the system. Kodura and Weinerowska [4] investigated the difficulties that may arise in modeling of water hammer phenomenon. Wood [5] compared MOC and Wave characteristics Method (WCM) showing that for the same modeling accuracy, the WCM requires less execution time. Afshar and

Rohani [6] proposed an implicit Method of Characteristics in order to alleviate the shortcomings and limitations of the mostly used conventional Method of Characteristics (MOC). Fouzi and Ali [7] used WHAMO program and AFT Impulse software for investigation of wave propagation in the case of sudden closure of valves in gravity adduction pipes. In this paper AFT Impulse software package that is based on method of characteristics, was used for numerical solving the water hammer equations. Obtained results for non-stationary and transient flow regimes are used to analyze the operation of the plant and improve its functioning in terms of reliability and avoidance of unstable and unsteady work regimes.

2. Hydropower Plant Bistrica Hydropower Plant Bistrica is a impoundment – type plant located on the river Uvac. The water to the turbine is fed through the tunnel and two penstocks. Penstocks transport water up to two Francis turbines. The average gross head of the plant is 376 m, the installed flow Q IN = 2x 8 =36 m3/s, installed power N IN = 2x52 = 104 MW, and possible average annual production of electricity is 350 GWh [9].

Figure 1. Longitudinal section and details of the penstock of HPP Bistrica

The tunnel is a circular cross-section with a diameter 4 m and length 8026 m. At the entrance of the tunnel is a inlet gate valve is mounted. Pressure in the tunnel ranges from 1,72 bar at the entrance to 4,66 bar at the end of the tunnel. The water velocity in the tunnel is v = 2,86 m/s and the slope of the tunnel is 2‰. At the end of the tunnel is a differential surge tank that is assigned to receive excess water from the penstock during unsteady working regimes. At the end of the tunnel is a valve house in which the butterfly valves are placed for each of the two penstocks. The penstocks are with variable diameter (from 2,20 m to 2 m), lengths 1357 m, and supply water to two Francis turbines with vertical shafts, located in the power house of HPP Bistrica. The maximum allowed pressure in penstocks is 43,8 bar [8]. The configuration and dimensions of the penstocks are given in Figure 1.

3. Numerical flow simulation in the penstock Numerical flow simulations were conducted in the case of non-stationary fluid flow through HPP tunnel and penstock in AFT Impulse software package. Non-stationary flow occurs as a result of rapid closure of the

main inlet turbine valve. Main inlet turbine butterfly valve with a diameter of the DN2000 mm is located right in front of the turbine. The simulations were conducted for different closing times of the main inlet valve. The closing valve characteristic is defined by the curve describing the change of the coefficient of flow Cv during the closing time and in the AFT Impulse program is marked with Cv. Figure 2 schematically shows HPP Bistrica tunnel and penstock with reinforcement and markings from the program AFT Impulse.

Figure 2. Schematic view of tunnel and penstock of HPP Bistrica from the AFT Impulse program

During the stationary flow of water through HPP Bistrica, does not come to the rising of pressure in the penstock, so there is no risk of damaging the penstock [9]. For this operating regime, numerical simulations are not given in this paper. Numerical simulations of non-stationary flow were performed for the operation of the HPP with total flow Q = 36 m3/s, with reduced flow and for operation of HPP with one turbine. All numerical simulations were performed for two cases of closing the main turbine inlet valve, with closing time t = 80 s (fast closing) and time t = 150 s (slow closing).

4. Numerical simulations results 4.1 Hydropower plant operation with total flow, Q = 36 m3/s I case – closing time of main turbine inlet valve t = 80 s : The curve of the flow coefficient change during the closing of main turbine inlet valve is shown in Figure 3.

Figure 3. Main turbine inlet valve closing curve

Figure 4. Change of pressure versus time in penstock

After numerical simulations are performed, curves of pressure and flow change versus time in penstock (Figure 4 and 5) are obtained, as well as surge tank surface gradeline versus time (Figures 6 and 7).

Figure 5. Change of flow versus time in penstock

Figure 6. Surge tank surface elevation change versus time

Figure 7. Flow through surge tank versus time

II case – closing time of main turbine inlet valve t = 150 s :

The curve of the main inlet valve coefficient of flow change during the closing is shown in Figure 8. After numerical simulations are performed, curves of pressure and flow change versus time in penstock (Figure 9 and 10) are obtained, as well as surge tank surface gradeline versus time (Figure 11).

Figure 8. Main turbine inlet valve closing curve



Figure 11. Surge tank surface elevation change versus time

Figure 12. Flow through surge tank versus time

On Figure 4 and 9 can be seen that decreasing of closing time of main inlet valve cause the pressure incerase in penstock. Maximum pressure in first case (t = 80 s) is 50 bar and it appears 65 s after total closure of main turbine inlet valve. Maximum pressure in second case (t = 150 s) is 41 bar and it appears 25 s after total closure of main turbine inlet valve. In both cases, the surge tank on HPP can take excess amount of water from the system that occurs during non-stationary regime of operation [10]. Since this is differential surge tank narrower cylinder takes 512,85 m3, up to the level of 819 m (Figure 6 and 10) while the remaining amount of water overflows in wider cylinder up to the level of 804,25 m. Filling level of the wider cylinder is obtained by calculating the volume of water that overflows from the narrower to the wider cylinder. When HPP operates with total flow (Q = 36 m3/s), the wider cylinder receives the largest amount of water, and only this filling level is given in the paper as the highest.

4.2 Hydropower plant operation with one turbine I case – closing time of main turbine inlet valve t = 80 s :

The curve of main turbine inlet valve closing is same as in first considered case, shown in Figure 3. After numerical simulations are performed, curves of pressure and flow change versus time in penstock (Figure 13 and 14) are obtained.


II case – closing time of main turbine inlet valve t = 150 s :

The curve of the change of the coefficient of flow during the closing of the pre-turbine gate valve is shown in Figure 7. After numerical simulations are performed, curves of pressure and flow change versus time in penstock (Figure 15 and 16) are obtained.




When HPP operates with one turbine decreasing of closing time of the main turbine inlet valve causes the incerase pressure in penstock. Maximum pressure in first case (t = 80 s) is 51 bar and it appears 65 s after total closure of valve (Figure 13). Maximum pressure in second case (t = 150 s) is 41 bar and it appears 5 s after total closure of valve (Figure 15). Numerical simulations were conducted also in regimes when HPP operates with reduced flows and with the same closing times of the main turbine inlet valve (t = 80 s and t = 150 s). For both cases, reducing the flow to one third Q 1/3 = 12 m3/s causes the system to work safely. Maximum pressure that occurs in this case is lower than maximum allowed pressure in penstock.

5. Conclusion Numerical simulations show that unsteady flow and transient regimes in the HPP Bistrica caused by different closing times of the main turbine inlet valve cause the changes in pressure and flow in the penstock over time. Since the maximum allowed pressure in the pipeline is 43,8 bar, it can be concluded that when HPP operates with maximum instaled capacity and closing time t = 80 s of the main turbine inlet valve, there is a risk of damaging the penstock because the maximum pressure in this case is 50 bar. When the HPP operates with total flow and closing time t = 150 s, there is no risk of damaging the pipeline because the maximum pressure in this case is 41 bar, and the vibrations caused by pressure change after the turbine fall from the grid is significantly lower. When HPP operates with one turbine and valve closing time t = 80 s, there is also a risk of damaging the penstock because and increased vibrations in hydropower plant. The maximum pressure in this case is 51 bar. When the HPP operates with one turbine and closing time t = 150 s, there is no risk of damaging the pipeline because the maximum pressure in this case is 41 bar.

Acknowledgements: Paper is result of project TR33040, Revitalization of existing and designing new micro and mini hydropower plants (100 to 1000 kW) in the territory of South and Southeast Serbia, supported by the Ministry of Education, Science and Technological Development of the Republic of Serbia.

References [1] [2] [3] [4]

Asli, K.H., Naghiyev, F.B.O., Haghi, A.K., Some aspects of physical and numerical modeling of water hammer in pipelines, Nonlinear Dynamics, 60 (2010), pp. 677–701 Izquierdo, J., Iglesias, P.L., Mathematical Modelling of Hydraulic Transients in Simple Systems, Mathematical and Computer Modelling, 35 (2002), pp. 801-812 Filion, Y.R., Karney, B.W., Sources of error in network modeling: a question of perspective, Journal AWWA, 95:2 (2003), pp. 119-130 Kodura, A., Weinerowska K., Some Aspects of Physical and Numerical Modeling of Water Hammer in Pipelines, International Symposium on Water Management and Hydraulic Engineering, Ottenstien, Austria, 2005

[5]

Wood, D.J., Waterhammer Analysis - Essential and Easy and Efficient, Journal of Environmental Engineering, 131:8 (2005), pp. 1123–1131 [6] Afshar, M.H., Rohani, M., Water hammer simulation by implicit method of characteristic, International Journal of Pressure Vessels and Piping, 85 (2008). pp. 851–859 [7] Fouzi, A., Ali, A., Comparative Study of the Phenomenon of Propagation of Elastic Waves in Conduits, Proceedings of the World Congress on Engineering, London, U.K., 2011 [8] Hydropower Plant Kokin Brod – Bistrica, A monograph on the construction of the first in a row of Lims Hydropower plants, Lims HPP, Nova Varoš, 1967. [9] Spasic, J., Water hammer in Turbine Plants, Master thesis, University of Nis, Nis, 2016 [10] Ivetic, V. M., Flow in pipes, Beograd, 1996



Optimization of theMeteorogical Guyed Mast MiladaPezoa(CA), VukmanBakićb, Lato Pezoc Laboratory of Thermal Engineering and Energy, Institute of Nuclear Sciences VINČA, University of Belgrade, Belgrade, Serbia,RS, [email protected] b Laboratory of Thermal Engineering and Energy, Institute of Nuclear Sciences VINČA, University of Belgrade, Belgrade, Serbia,RS, [email protected] c Institute of General and Physical Chemistry,University of Belgrade, Studentski Trg 12-16, 11000 Belgrade, Serbia, RS, [email protected] a

Abstract: Geometrical and working parameters of the guyed masts have been defined in the European standards EN 1993-3-1:2006 and EN 1993-1-6:2007 ensuring local and global stability and serviceability of the structure. FEM (Finite Element Method) was used for structural analysis, modal analysis and buckling analysis of the guyed mast exposed to wind action. The investigated mast is a 139m tall tower with support guys, consisting of standard beam elements. For the given height, geometrical in-plan and bracing schemes of the mast, wind loading and material data, the optimization procedure is performed for the dimensions of all mast elements, attachment points of the guys, the distances of the guys’ foundations from the mast, etc. The aim of the article is to optimize the tall guyed masts, seeking the minimal mast weight, minimal displacement of the top of the mast and minimal values of natural frequencies. Keywords: Optimization; Guyed mast; Standard score analysis.

1. Introduction The meteorological guyed masts are widely used for feasibility studies of possibility of installing wind farms at the appropriate locations. The stability of the mast structure is jeopardized by wind load, ice load and earthquake. Stability of the mast can be analyzed according to structural analysis of lattice structures and tall towers based on FEM (Finite Element Method). European standards EN 1993-3-1:2006 [1] and EN 1993-16:2007 [2] already provides some guidelines about the basis of design, structural analysis, ultimate and serviceability limit states and buckling of the components of masts and towers. But the construction of guyed masts can be improved taking into account safety and serviceability requirements. The parameters that can be optimized are: the material and dimensions of the beam elements used for leg members and bracing members, a number of guys, the position of the guys at the tower construction, the distance between the shaft and the connection point of guys and ground, the position of the connection between the guys and the shaft, the material and dimensions of the guys elements, the width of the mast in the cross-section etc. The optimization of the guyed masts has been performed in the previous research. Structural analysis is extremely important and complicated because of non-linear behavior, slenderness and compliant ‘guysupport’ system [3-10]. In this paper numerical simulations were performed based on finite element method and optimization of the geometry of the mast structures was performed for static wind loads according to the patch load method. Modal analysis was used to determine the frequency of the oscillations and mode shapes. The obtained values from these analyses are the values of maximal displacement and maximal stress and natural frequency of the oscillations of the mast. The obtained results have been subjected to analysis of variance (ANOVA) to show relations between parameters. In order to enable more comprehensive comparison between investigated cases, standard score (SS) has been introduced. Principal component analysis (PCA) has been applied to classify and discriminate analyzed cases.

2. Optimization Problem The optimization of the mast is performed for static analysis, and the static wind load is multiplied by the coefficient of turbulent loading according to the patch load method [11, 1]. In that manner, the static problem was solved, despite the fact that wind forces are of a dynamic nature. The static analysis is used to determine

stress/displacement field by using FEM. Modal analysis was used to determine the values of the natural frequency and mode shapes. This analysis was also obtained from numerical simulation based on FEM. In this paper, finite element method is used for modeling behavior of a guyed mast exposed to wind action. The discrete model was used for modeling cables, presenting every cable element as a series of several truss elements [12]. Numerical simulations of structural response were performed for static analysis according to the patch load method, providing the values of displacement and von Mises stress. Modal analysis was performed in order to obtain values of natural frequencies for first 12 mode shapes. The tower is a 139m tall guyed mast consisting of 46 segments (3m length each and one segment 1m length). Connections between elements are defined as rigid. The structure of the mast is shown in Fig. 1. The tower is supported by cables hooked for each peak of the triangle at appropriate height and has two positions on the ground. Each level is held by three guys with an angle of 120degrees between them. The cables are made of elastic material and they are modeled to be able to change shape and length.

Figure 1. Scheme of the guyed mast with four levels of guys.

Cables are modeled using a uniaxial tension-only finite element with three degrees of freedom at each node: translations in the nodal x, y and z directions. Cables are modeled by using so-called discrete modeling [12]. FEM is used for the optimization method to determine displacement/stress field of the mast for structural analyses. The initial data for mast optimization are as follows: • The height of the mast; • Disposition of the mast elements; • Maximal displacement of the top of the mast, according to EU standards; • Maximal allowed stress, according to EU standards; • Wind loading according to patch method; • The material of the beam elements used for tower construction; • The material of the ropes used for guys support. These parameters are constant and fixed in all cases used for optimization problem. The parameters which are variable are as follows: • The diameter of the beam elements used as leg elements; • A number of guys used to support the tower; • Geometry of the supported elements of the tower. These parameters are varied and appropriate choice of their values can optimize the mast in the manner to provide better characteristics and lower price of the mast construction. The optimization problem is formulated as follows:

Y1 = min {mass of the mast} Y2 = min {displacement of the top of the mast} Y3 = min {von Misses stress}

(1)

Y4 = min {natural frequency} The mass of the mast is one of the most important parameters of the mast. The geometry of the beam elements can influence the mast mass and the price of the whole construction. Guyed masts are used in telecommunication industry and it is very important that the disposal of the top of the mast is minimal to provide better quality of the telecommunication signal. Stability of the mast depends on the stress in the beam elements. This parameter has a strong influence on the stability of the construction. The natural frequency must be controlled, because of the insuring stability of the construction.

3. Finite Element Model The tower structure is modeled using quadratic two-node finite strain beam elements with six degrees of freedom with an internal node in the interpolation scheme. Cables are modeled using a uniaxial tension-only finite element with three degrees of freedom at each node: translations in the nodal x, y and z directions. In order to ensure the stability of the mast, it was necessary to impose some constraints concerning the structure. The constraints are introduced according to Eurocodes and include the local and global stability constraints, limitations on the slenderness in mast elements, and strength constraints. The strength constraints are checked for all structural elements of the mast, accounting for both the axial force N and bending moment M for the allowable stresses σ that depend on the diameters of the elements:

σ=

N M ± ≤σ A W

(2)

The stability of the leg and bracing elements, according to Eurocode 3, Part 3-1, is insured if the inequation holds:

N ≤

χ Af y γ M1

(3)

The safety assessment for fatigue should be carried out inaccordance with [13], using:

λ ∆σ E ∆σ E 2 =

(4)

where λis the equivalence factor to transfer ∆ σ E to N c = 2 x 106 cycles 1

 N m λ= 6   2 ×10 

(5)

∆ σ E is the stress range associated to N cycles. The lateral displacement D of the top of the mast is constrained to:

D≤ where h is the height of the mast.

h 100

(6)

4. Statistical Analysis The optimization was performed by using principal component analysis (PCA). This method was proved to be successful to classify and discriminate the different cases, and to discover the possible correlations among calculated parameters. The evaluation of ANOVA and the PCA analyses of the obtained results were done by StatSoft Statistica 10.0® software [13]. The response surface regression method (RSM) was selected to estimate the main effect of the process variables on guyed mast stability. The calculated data used for this study were obtained using a central composite full factorial design (3 level-3 parameter) with 27 runs (1 block). The independent variables were: n– number of guys; d 1 – main beam dimensions [m], d 2 – dimensions of the crossbeam [m], m– total mass [kg], D – displacement of the top of the mast [m], V– maximal von Misses stress [N], F – frequency [Hz], SS – standard score. A model was fitted to the response surface generated by the numerical results. The model was obtained for each dependent variable where factors were rejected when their significance level was less than p

Development and Validation

Development and Validation

Suggest Documents

DEVELOPMENT AND CONVERGENT VALIDATION

Method Development and Validation of

Development and Validation of UV

Development and Validation of Objective

Development and validation of spectrophotometric

Development and Validation of Spectrophotometric

Development and validation of UV

Development and Preliminary Validation of

HPTLC METHOD DEVELOPMENT AND VALIDATION

Development and validation of Coaches

Development and Validation of Loperamide

Development and validation of AIMFREE

Analytical Method Development and Validation

hplc method development and validation

DEVELOPMENT AND VALIDATION OF A

DEVELOPMENT AND VALIDATION OF AN

Development and validation of a nutritional and

development and validation of method for olanzapine in bu validation ...

Model Development and Validation Governance and Standards

DEVELOPMENT AND VALIDATION OF A SENSITIVE AND

Development and V Development and Validation of ...

THE DEVELOPMENT AND VALIDATION OF A ... - OhioLINK ETDhttps://www.researchgate.net/.../THE-DEVELOPMENT-AND-VALIDATION-OF-A-NO...

KTH Engineering Sciences Development, Validation and Application ...

Development and Validation of Credit% Scoring Models