Fuel Processing Technology 166 (2017) 258–268

Contents lists available at ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

Research article

Combustion chamber design and performance for micro gas

turbine application
Ibrahim I. Enagi a, K.A. Al-attab b, Z.A. Zainal c,⁎
a
Department of Mechanical Engineering, School of Engineering Technology, Federal Polytechnic, P.M.B. 55, Bida, Niger State, Nigeria
b
Department of Mechanical Engineering, Faculty of Engineering, Sana'a University, Sana'a, Yemen
c
School of Mechanical Engineering, Universiti Sains Malaysia, Engineering Campus, 14300 Nibong Tebal, Penang, Malaysia

a r t i c l e i n f o a b s t r a c t

Article history: Micro-gas turbines (MGT) are small-scale independent and reliable distributed generation systems that offer po-
Received 23 January 2017 tential for saving energy and reducing carbon monoxide (CO) emissions. They are expected to play a vital role in
Received in revised form 31 May 2017 future energy supplies for remote locations with or without grid connections. In this paper, a design and devel-
Accepted 31 May 2017
opment of a combustion chamber for micro-gas turbine was performed by SOLID-WORKS and computational
Available online 19 June 2017
fluid dynamics (CFD) ANSYS-FLUENT simulation software. Different chamber geometries were used to simulate
Keywords:
with species transport and non-premixed combustion models to determine the optimum chamber design. The
Micro gas turbine best chamber geometry adopted after optimization was 50 mm flame holder diameter, 60 cm chamber height,
Turbocharger having 4 holes of 6, 8 and10 mm with dead zone between the combustion zone and dilution zone. A two-stage
Combustion chamber MGT was developed based on vehicular turbochargers to test the chamber. The experimental test of the chamber
Flame holder with liquefied petroleum gas (LPG) fuel resulted in a stable combustion with CO emission below 100 ppm and
ANSYS-FLUENT turbine inlet temperature below 900 °C.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction turbine rotating at 3000 RPM connected to a conventional generator

via gear box in a split- shaft designed whereas the compressor speed
Micro-gas turbine (MGT) usually produces between 25 and 500 kW varies with output load resulting in a better part load efficiency [6,7].
of electrical power, it has minimal maintenance and operational cost, The developmental work in small stationary and automotive gas-
high power density and low emission. An important factor that attracts turbines were initiated by automotive industries in 1950′s and served
researchers to develop MGT's especially for renewable energy fuel types as fundamental achievement in today's MGT technology [8]. Modern
is that it can be operated with various kinds of fuels [1]. Micro-gas tur- MGT combines the reliability of an aircraft generator with low cost au-
bine in recent times is given attention for decentralised generation of re- tomotive turbocharger [8]. The common technique used to increase
newable energy [2]. There has been a renewed interest on the MGT the power density of internal combustion engines is turbocharging
development and deployment on small scale distributed cogeneration which reduces pollutant emission and fuel consumption [9]. The overall
(DG) and poly-generation concepts [3]. When compared with other turbocharger performance and turbine power have significant effects
technologies for small scale power generation, they offer numerous ad- on the thermal energy transfer and engine volumetric efficiency [10,11].
vantages which include, high grade heat and lower emission levels, The design of a combustion chamber based on temperature homo-
compact size, reduced noise and vibrations, easy operations and instal- geneity and CO emission is a critical issue in the development of the
lations [4]. MGT. Although, some typical design and experimental studies on differ-
MGTs are essentially of two types, the first is composed of a high ent micro combustor configurations were carried out in the past, there
speed single shaft unit carrying centrifugal compressor and radial tur- is need for further design and development to improve on fundamental
bine (ranges from 50,000 RPM to 120,000 RPM) on the same shaft as issues such as low outlet temperature and CO emissions. Flame sustain-
an electrical synchronous machine. In this design, the compressor ability over a range of operating mass flows and air-fuel (AF) ratios in a
speed remains constant at generator rated speed resulting in a signifi- high power density micro combustion were investigated [12,13]. A high
cant drop in efficiency at part load [5]. The second type uses a power and uniform temperature distribution along the wall of the micro com-
bustor flame tube was achieved using a stainless steel based combustor
⁎ Corresponding author. [14,15]. Another micro-axial stainless steel based combustor was devel-
E-mail address: mezainal@usm.my (Z.A. Zainal). oped to improve thermal performances of MGT [16,17].

http://dx.doi.org/10.1016/j.fuproc.2017.05.037
0378-3820/© 2017 Elsevier B.V. All rights reserved.
 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268 259

The applications of simulation software have rapidly increased over determined based on Garrett GT25 turbocharger specifications and
the last decade in improving many complex processes including com- compressor maps. Maximum air flow rate of 0.15 kg/s at 1.4 barg was
bustion chambers. Computational fluid dynamics (CFD) has been exten- considered to achieve compressor efficiency of about 70%. Optimum
sively used as an important design tool of combustion chambers. chamber should achieve a complete combustion with low CO emissions
Numerous studies have utilized CFD simulation for the design and opti- and stable flame propagation inside the flame tube in a compact form.
mization of burners [18,19] and combustion chambers in MGT [6,20– Table 1 shows the main boundary conditions for air and fuel inlets,
22], aviation and industrial gas turbines [17,23]. chamber outlet and walls.
In the present work, a combustion chamber was designed and devel-
oped using CFD ANSYS-FLUENT simulation software. Species transport 2.2.1. Species transport model
and non-premixed combustion models were used to determine the op- For flow analysis, k-epsilon viscosity model was used since it is rec-
timum flame holder chamber geometries. The designed combustion ommended by the FLUENT theory guide for laminar-turbulent transi-
chamber was investigated through optimization process and the effects tion and turbulent flow ranges. Reynolds stress model will be used in
of different variables and geometries on the combustion stability were the future to simulate different type of commercial diesel swirling injec-
studied. Combustion stability was also verified experimentally during tors to be compared with the experimental test with diesel fuel. As for
the initial tests of a two-staged turbocharger based MGT. the combustion, energy equations were activated (i.e. non-adiabatic
combustion) with species transport model used first due to its lower
2. Combustion chamber design computing demand compared to the non-premixed combustion
model. This step will provide the initial estimation of the minimum
For chamber design, a criterion for selecting suitable combustor ge- size in which the combustion will start taking place. Diesel-air and pro-
ometry was carefully examined followed by designed calculation of the pane-air mixture were tested separately at 20% excess air with 0.15 kg/s
dimensions. The combustion chamber design is faced with inherent air flow rate. And with such high flow, turbulent combustion is expected
challenges such as emission control, chamber materials temperature in small geometries while larger volumes will experience laminar flows
limitations and flame stability. The steps taken in designing the cham- in some parts of the chamber. Therefore, both laminar Finite-Rate and
ber are: SOLIDWORKS drawing of the chamber, and simulation using the turbulent Eddy-Dissipation equations were used in the volumetric
ANSYS-FLUENT software. combustion model. Main geometry parameters studied with this
model were the chamber height and flame tube diameter. Other vari-
2.1. SOLIDWORKS drawing of the chamber ables such as chamber diameter and flame tube holes configuration
were neutralized. For chamber diameter, air jacket width of 25 mm
The combustion chamber consists of the air jacket and flame tube was fixed, thus, making this variable dependent on the flame tube diam-
with 100 mm fixed height conical exit of the chamber. Chamber inlet eter variable. As for the flame tube configuration, 8 holes of 6 mm diam-
and outlet diameters were also fixed at 50 mm diameter to match the eter were used in each row with 10 mm distance between the rows.
specifications of the first stage MGT Garrett GT25. Main manipulated ge- Thus, number of rows will increase accordingly with the chamber
ometries were: chamber height, chamber diameter, flame tube diame- height. Geometries with sustainable combustion were then further op-
ter and combustion zones configurations on the flame tube. Flame timized using radiation and non-premixed combustion models.
tube diameter was varied in the range of 50–150 mm, thus, chamber di-
ameter was varied accordingly to leave air jacket space of 25 mm, while 2.2.2. Non-premixed model
chamber height was varied from 300 to 1000 mm. Simple flame tube To achieve stabilized combustion in MGT combustors, an improved
geometry with fixed rows of holes of 6 mm diameter through the tube understanding of turbulent combustion through CFD simulation is re-
height was studied first. However, for the final flame tube optimization, quired. Moreover, the accuracy of this simulation is highly dependent
three zones were identified on the flame tube: premix zone, primary on the turbulence and combustion models. The non-premixed combus-
combustion zone and dilution or cooling zone. Each zone has a number tion model which employs the infinitely fast chemistry assumptions
of rows with a certain number of holes on teach row with space be-
tween zones. The number of rows was varied from 4 to 15 for each
zone and number of holes in each row was also varied from 4 to 10.
Two holes diameters arrangements in respective to the zones were Table 1
studied: 6, 8, 10 mm and 8, 10, 12 mm. Parameters set out in boundary conditions.
In order to simplify the meshing and simulation processes, flame Parameters Values
tube was built as a void inside the chamber instead of a solid body.
Fuel inlet
This was achieved by creating two separate bodies (chamber and
Temperature 300 K
flame tube) then subtracting the flame tube body from the chamber Pressure 2 barg
using Combine: Subtract feature in SOLIDWORKS. Mass flow rate (LPG) (20%–70% excess air) 0.0053–0.0075 kg/s
Mass flow rate (Diesel) (20%–70% excess air) 0.0056–0.0078 kg/s
2.2. Simulation using ANSYS-FLUENT software Air inlet
Temperature 530 K
Combustion chamber geometry was optimized using ANSYS 16.1 Pressure 1.4 barg
CFD simulations program. In the Work-bench, chamber geometry was Mass flow rate 0.15 kg/s

imported from SOLIDWORKS to the Design-Modeler tool. Meshing Outlet

(ANSYS ICEM CFD) tool was used to create the mesh and test its quality. Pressure 1.4 barg
After that, 3D simulation was performed on the mesh using Fluent CFD Back flow temperature 600 K

tool. Species transport combustion model was tested first followed by Inner walls
the non-premixed combustion model. The chamber is aimed at testing Materials Steel
liquid and gaseous fuels with MGT, thus, LPG (butane/propane) and die- Emissivity 0.5

sel (C12H23) fuels were tested. For this initial design simulation, no Outer walls
swirlers were used. However, different type of commercial diesel injec- Materials Steel
tors/swirlers will be simulated and compared to the experimental test Wall thickness 6 mm
Heat fluxes −10,800 W/m2
with liquid fuels in future work. Air inlet boundary conditions were
260 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268

was therefore applied to predict the diffusion flames. Further optimiza- 3. Experiment test rig
tion was carried out on the accepted geometries using non-premixed
combustion model. Probability density function (PDF) of diesel and 3.1. Setup
LPG fuels were used separately for the combustion model. LPG fuel in
Malaysia contains about 70% butane, 29.5% propane and 0.5% pentane/ A test rig is aimed at testing gaseous and liquid fuels on MGT was de-
other components in volume [24]. Therefore similar values were set in veloped. The test rig consists of a combustion chamber designed using
the PDF in order to compare the results with the experimental data. Ra- CFD simulation, two-stage turbocharger based MGT and exhaust gas
diation P1 model was introduced to take account of radiation heat trans- heat recovery unit (HRU). The optimum combustion chamber design
fer inside the chamber with stainless steel wall emissivity of 0.5. was fabricated using 6 mm thick stainless steel (SS) pipes for the
Convection heat flux loses through the outer jacket walls were found outer air jacket and flame tube, while the top plate was fabricated
to be about10800W/m2. The XY plots of temperature and CO for cham- using 10 mm thick SS plate. The optimum geometry is 600 mm in height
ber outlet were generated to calculate average turbine inlet tempera- and 152 mm inner diameter with a 100 mm height nozzle at the cham-
ture (TIT) and mole fraction of CO in ppm. Three main simulation ber exit. The flame holder was based on flame tube design of 690 mm in
outputs were investigated to determine the optimum design: height and 50 mm inner diameter. Combustion zones contained 4 rows
of holes with 8 holes in each row with largest dead zone of 350 mm be-
• CO emissions that indicated the combustion completion. tween primary and dilution zones as shown in Fig. 1(a). Holes diameters
• Outlet temperature that indicates the effectiveness of the dilution or for premixing, primary and dilution zones were 6, 8 and 10 mm, respec-
cooling zone. Hence, lower temperature is desired to prevent damag- tively. A 6 mm diameter LPG injector of was fabricated for the experi-
ing MGT turbine blades. ment based on the simulation. As for diesel injector, four sizes of
• Flame propagation inside the flame tube. The cases with disturbed commercial injectors 1–4 GPH with 30° spray angle and swirler will
flame where flame exits the tube to the air jacket were rejected be simulated and compared to the experimental test with diesel in a fu-
since the tube becomes surrounded by flame at some points without ture work.
air cooling causing damage to the tube material. Flame shape inside Garrett GT 25 was used on the first stage and Holset H1C for the sec-
the tube was also studied to prevent the direct contact between inten- ond stage. The compressor was removed from the H1C and replaced
sive and high temperature flames and tube walls. with aluminium friction disc connected to a Teflon brake dynamometer.
Exhaust gas from the second stage turbine was recovered in HRU heat
exchanger for hot air production. The HRU comprises 100 annular coun-
The main geometry manipulated variables were: chamber height, ter-flow tubes with 6 mm diameter cupper inner tube for gas flow and
flame tube diameter (chamber diameter will change accordingly) and 19 mm diameter mild steel outer tube for air flow. HRU design and test
hole zones on the flame tube. Zone arrangement on the flame tube in- performance were published in earlier work [6,25]. Other auxiliary units
cluded holes diameters, number of holes in one row, number of rows include, oil pump, oil cooling radiator, liquid fuels pump, MGT start-up
per zone and distance between zones which will be referred to as air blower and high volt DC transformer for spark plug ignition. Fig.
dead zones since it does not include any holes. 1(b) shows a schematic drawing for the system.

Fig. 1. (a) Combustion chamber drawing; (b) schematic drawing of the test rig.
 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268 261

3.2. Procedures and measuring equipment 4.2. Chamber optimization

For the initial test of the rig, LPG fuel was used to evaluate the com- The combustion chamber was further optimized with the radiation
bustion chamber, MGT and the Teflon brake dynamometer. The opti- P1 model and non-premixed combustion model with PDF for diesel
mum combustion chamber design was evaluated in terms of and LPG. Chamber heights were varied from 300 mm up to 700 mm
combustion stability, emissions and the capability of cooling zone to with flame tube diameter in the range of 50–100 mm. Simulation results
drop TIT down in the range of 900–1000 °C to prevent damaging turbine with LPG fuel will be illustrated in this section for better comparison
blades. Variety of liquid bio-fuels such as biodiesel, diesel-cooking oil with the experimental work.
blends and pyrolysis bio-oil will eventually be tested on the rig.
Five type-K thermocouples were used to capture the temperature
profile for the combustor, MGT and HRU. Also, a 2 barg pressure gauge 4.2.1. Effect of number of rows
was used to measure the MGT compressor pressure additional to four For the flame tube zones optimization, first arrangement of 6, 8,
differential pressure gages to measure pressure drop in the combustor, 10 mm hole diameters for premixed, combustion and dilution zones, re-
first and second stage turbines and HRU. Mechanical shaft power of the spectively, was fixed while changing other variables. First variable was
second-stage turbine or the power turbine was determined using number of rows were 4, 7 and 15 rows in each zone were tested with
50 mm Teflon brake disc connected to a 5 N spring scale to measure chamber heights in the range of 300 mm to 700 mm with 50–100 mm
the torque. Power turbine speed was limited to 20,000 RPM to prevent flame tube diameter range. It was noticed that increasing number of
overheating the Teflon brake disc and torque readings were taken at rows increases the flame bleeding out to the air jacked in all cases
10,000, 15,000 and 20,000 RPM. A vortex-tube cooler was used to cool with no noticeable difference between LPG and diesel. An example tem-
the disc and a 100,000 RPM optical tachometer was used to measure perature contours for 700 mm chamber height with 15 rows is shown in
the power turbine speed. Air flow rate was measured using hot-wire an- Fig. 2(a) and 7 rows for 300 mm chamber height in Fig. 2(b). On the
emometer while LPG flow rate was measure using a 100 LPM rotameter. other hand, in the 4 rows design, flame was mostly contained inside
For MGT start-up, a ring-type 750 W air blower was used. LPG was ignit- the tube even for the low chamber height (300 mm) with minor bleed-
ed with high-volt spark plug and at 0.2 barg compressor pressure, the ing at the rich fuel premix zone as shown in Fig. 2(c) and (d) for 700 mm
first-stage MGT can run in a self-sustain mode and the air blower can and 300 mm chamber heights respectively. Although the number of
be shut down. The MGT was tested in the range of 0.2–0.7 barg and rows has affected the flame pattern noticeably, it did not affect combus-
the measurement of air and LPG flow rates, temperature and pressure tion quality in term of maximum temperature and CO emissions.
profiles and shaft brake power were taken every 0.1 barg intervals.

4.2.2. Effect of holes diameters

4. Simulation results and discussions The effect of flame tube holes diameters on flame stability was inves-
tigated by increasing the diameters from 6, 8, 10 mm to 8, 10, 12 mm
The species transport model revealed the preliminary result of com- hole diameters arrangement as shown in Fig. 2(e) and (f). It was con-
bustion for different flame tube heights and diameters. However, it did cluded that increasing the cross flow surface area of the holes between
not indicate combustion completion since the model does not provide air jacket and combustion zone will lead to flame bleeding to the jacked.
CO emissions. Hence, non-premixed combustion model with radiation Hence, the increment in holes diameters resulted in a similar result to
model and convective heat losses from the chamber were used to deter- the addition of more rows. Another note is that the flame bleeding oc-
mine optimum chamber geometry. curred mostly at the rich flame zone which is expected since the
flame tends to propagate towards richer oxygen medium if the path
has low flow resistance. Thus 4 rows of 6 mm diameter holes were suf-
4.1. Species transport model ficient to maintain rich fuel flame inside the tube.

The simulations were first performed on different 3D geometries of

chamber heights and flame holder diameters using species transport 4.2.3. Effect of flame tube diameter
combustion model. For air flow rate of 0.15 kg/s, chambers with low Chambers with low height below 600 mm did not show a good per-
height below 300 mm were not able to sustain flame at all and even formance in terms of flame stability and CO emissions. Therefore, three
with the patched ignition temperature of 2000 K, temperature dropped chamber heights of 600, 700 and 800 mm were further compared with
directly back to the boundary temperature. Table 2 shows the combus- 25, 75 and 100 mm flame tube diameters. Results shown here in this
tion sustainability with diesel and propane fuels for chamber heights in section are for LPG PDF simulation. However, diesel fuel has shown sim-
the range of 300–700 mm. Propane fuel has proven to be easier to ignite ilar trends with the different geometries, but with slightly higher tem-
and sustain the flame at smaller geometries. As for larger geometries, perature and CO emissions. In order to determine the optimum
combustion was easily sustainable, however, minimal volume and sur- geometry, the main manipulated parameters were the average CO
face area is preferable to reduce heat loses and manufacturing cost. emissions and temperature at the chamber outlet. Fig. 3(a) and (b)
show average CO emissions and temperature, respectively, for the
three heights with LPG fuel and 70% excess air. It can be noticed that
most of the geometries generated outlet temperature slightly above
Table 2 1200 °C, hence, temperature was not a considerable variable for the
Results obtained for species transport model. comparison between geometries. On the other hand, flame tube diame-
ter showed a considerable effect on CO emissions. Although, the flame
Chamber height (mm) 300 400 500 600 700
tube diameter had no effect on combustion setup in terms of air-fuel
Flame holder diameter (mm)
(AF) ratio, it affected the velocity and turbulence profiles inside the
50 P P P,D P,D
75 P,D P,D P,D P,D tube. Therefore, the smaller 50 mm diameter flame tube with higher
100 P P,D P,D P,D P,D flame turbulence and better air-fuel mixing inside the tube resulted in
125 P,D P,D P,D P,D P,D a slight enhancement in combustion reducing CO emissions in the
150 P,D P,D P,D P,D P,D range of 100–150 ppm compared to 200–300 ppm for larger flame
Note: (P) sustainable combustion with propane; (D) sustainable combustion with diesel. tubes.
262 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268

Fig. 2. Temperature contours for (6,8,10) mm configuration (a) 700 mm height with 15 rows; (b) 300 mm height with 7 rows; (c) 700 height with 4 rows (d) 300 mm height with 4 rows;
(e) 700 mm heights with 4 rows (8,10,12) mm configuration (f) 300 mm height with 4 rows (8,10,12) mm configuration, all with LPG fuel.
 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268 263

Fig. 3. (a) Average CO mole fraction for different geometries; (b) average outlet temperature for different geometries; and Average CO mole fraction for: (c) 800 mm height with and
without radiation model, (d) 60 & 70% excess air, (e) dead zones A and B, all with LPG fuel.

4.2.4. Effect of radiation model 4.2.5. Effect of excess air supplied

Internal heat transfer inside the flame tube is mainly carried out by In general, turbine engines require much higher amount of combus-
the forced convection by flame propagation and also natural convection tion excessive air compared to other thermal engines. Only a fraction of
causing heat to travel vertically to the top of the chamber. However, all the supplied air is used for a near-stoichiometric combustion and the
air and fuel species as well as the internal surfaces of the flame tube are rest is used to dilute and cool down the combustion product gasses to
also heated by the radiation emitted from the flame. Radiation has a prevent damaging of the turbine blades. Also, the amount of excess air
positive effect on combustion through additional heat to further burn is a major parameter in the thermal efficiency of a combined heat and
the remaining unburnt CO species. Fig. 3(c) shows the effect of P1 radi- power systems since most of the fuel input power is converted to sensi-
ation model on the average CO emissions at the chamber outlet. ble heat at the turbine exit [26]. In previous studies [27,6] a Mitsubishi
Flame tube diameter of 50 mm showed a superior performance TD05 turbocharger which has similar capacity to the GT25 turbocharger
compared to the higher diameter of 75 mm. Also, extending the height used here, was tested at nearly maximum flow capacity. Stable MGT op-
of the chamber up to 800 mm with the additional heat loses as surface eration was achieved using LPG fuel with excess air supply in the range
area increases and the additional material cost did not show any com- of 30–70%. In this simulation, air flow rate was fixed at 0.15 kg/s while
bustion enhancement. Thus, further testes were carried out on 600 reducing fuel flow rate to achieve higher excess air values. The effect
and 700 mm heights with 2 in. flame tube diameter. of excess air on CO emissions is shown in Fig. 3(d) with 60% and 70%
264 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268

excess air. A considerable drop in average CO mole concentration at the configurations A and B shown in Fig. 3(e) were compared. Early intro-
outlet was achieved by reducing fuel flow rate while maintaining same duction of dilution air (configuration A) into the non-fully established
air flow which is expected. However, in practice, excessive amount of air flame resulted in a considerably higher CO emissions especially in the
without controlling and canalling air to the dilution zone will result in a limited chamber height of 600 mm as shown in Fig. 3(f). The length of
near lean-limit combustion. This can affect combustion stability signifi- the primary combustion zone in MGT combustion chambers is a major
cantly and lead to frequent flame blowout. concern in combustion quality [28]. Extending the primary zone (con-
figuration B) to allow most of the species to complete the reaction be-
4.2.6. Effect of dead zones fore introducing the dilution air has resulted in a significant drop in
It was established earlier in the preliminary optimization of the CO emissions for both 600 and 700 mm chamber heights.
zones geometry that a stable combustion can be achieved with 4 rows The 600 and 700 mm chamber heights did not show a considerable
of holes with 6, 8, 10 mm holes diameters. However, the distance be- difference in CO emissions and chamber outlet temperature. Thus, the
tween zones (i.e. dead zone) also plays a considerable role in flame sta- 600 mm chamber height, 50 mm flame tube diameter, with 4 rows of
bility and combustion quality, but to a lesser extent. Two dead zone holes of (6,8,10) mm diameters, and extended dead zone between

Fig. 4. Optimum chamber geometry (a) temperature plates; (b) CO mole fraction plates; (c) stainless steel flame tube visual inspection after test; and optimum geometry simulated at the
experimental conditions: (d) flame tube temperature, (e) temperature plates, all with LPG fuel.
 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268 265

combustion and dilution zones was chosen as the optimum chamber which surface oxidization did not take place. Combustion has extended
geometry. Using XY plot option in Fluent, average temperature and CO through the dead zone between combustion and dilution zones, while
mole fraction for the chamber outlet were 1218 °C and 102 ppm, respec- cold air jets through dilution holes has pushed the flame front away
tively. Fig. 4(a) and (b) show the temperature and CO distribution in from tube walls to the center of the tube. This indicated that the dilution
two vertical XY, ZY plates. The flame was well distributed within the zone geometry was sufficient to cool down the combustion product gas
flame holder with minor flame bleeding at the rich mixture mixing and drop the turbine inlet temperature below 900 °C. Simulation tem-
zone. perature profile results of the flame tube inner and outer walls and
chamber temperature contours are shown in Fig. 4(d) and (e). Air jacket
5. Combustor performance played a significant role in cooling the flame tube walls to prevent dam-
aging the tube. A comparison between the hot zones in simulation with
For the initial combustor and MGT test, LPG fuel was used since it the visual inspection of the flame tube showed a good agreement and
provides more flexibility and control on turbine speed and output explained the hot spot below the premixed zone cause by the flame
power. The non-premixed combustion CFD simulation was repeated leakage.
on the optimum geometry but at the new boundary conditions obtained The CFD model provided valuable data that were not attainable oth-
from the experiments with LPG fuel at maximum turbine power. erwise by experiment such as the flow pattern and AF ratios at each
Boundary conditions were set as 0.7 barg, 0.07 kg/s air flow rate and zone inside the chamber. AF ratios varied significantly from the center
37% excess air in order to verify the CFD model. Fuel species were set of the flame tube towards the tube walls in all zones. This is because
in the PDF at 70% butane and 30% propane to match the LPG used in air was introduced radially towards the center of the tube where fuel
the experiments. The flame tube was visually inspected after the test is concentrated resulting in a reduction in AF ratio towards the center.
as shown in Fig. 4(c). It was noticed the flame started to propagate At the premix zone, low AF ratio of 3 dominated most of the tube and
right after the premixed zone causing stainless steel surface oxidization only elevated at the air inlet holes on the tube walls. This resulted in a
and discoloration with bluish hot stains on the flame tube walls be- small combustion spots at the holes as shown by the temperature vec-
tween premix and dilution zones. On the other hand, the tip of the tors in Fig. 5(a) for the first row in premix zone. Also, the flow pattern
tube after the dilution zone holes maintained it metallic shining appear- shown in the same figure reveals a slight fuel leakage towards air jacked
ance indicating that the metal surface temperature was below 400 °C in caused by the air stream around the tube that cause eddies and lower

Fig. 5. Optimum geometry simulated at the experimental conditions with LPG fuel: (a–c) velocity vectors colored by total temperature for (a) premixed zone (b) combustion zone (c)
dilution zone; (d) mole fraction of CO; (e) formation rate of thermal NOx; (formation rate of prompt NOx).
266 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268

Fig. 6. (a) MGT temperature profile; (b) chamber CO and NOx emissions.

pressure zones where air stream is separated from the tube. The size of oxygen for the reaction. Prompt NOx, on the other hand, can occur at
the fuel rich zone at the center of the tube shrinks as fuel is further con- lower temperature when free hydrocarbon radicals react with nitrogen
sumed down the tube towards the combustion and dilution zones. As [29]. The generation of thermal NOx was concentrated mainly at the
for the combustion zone (rows 5–8) in Fig. 5(b), the fuel rich zone highest temperature zone near the inner wall of the flame tube shown
with AF ratio of 4 is easily identified by its lower temperature earlier in Fig. 4(e) where AF ratio is slightly above stoichiometric condi-
surrounded by flames in a rich to stoichiometric combustion (AF 10– tion. As for the prompt NOx, the generation started at the rich fuel pre-
15). The flow pattern in the combustion zone is more stable compared mix zone where fuel chains broke down to lighter hydrocarbon such as
to premix zone with uniform air flow through all the holes. Fig. 5(c) methane that can react with nitrogen to form NOx. The formation of
shows the flow pattern and temperature profile in the dilution zone. prompt NOx was also detected at the combustion zone but it did not
Cooling air is pushed through the larger holes of 10 mm in diameter
causing a significant drop in average temperature before the turbine.
Table 3
In general, the flame started as rich combustion hot spots near the air Comparison between experimental and CFD simulation results.
inlet holes in premixed zone with low AF ratio generating high amounts
of CO as shown Fig. 5(d). Rich combustion also propagated between the Parameters Experiment Simulation Error (%)

combustion and dilution zones as indicated by the high CO concentra- Fuel composition 70% butane and 29.5% 70% butane and 30% –
tion in the same figure. However, CO concentration dropped significant- (Vol%) propane propane
Air inlet pressure 0.7 0.7 –
ly at the dilution zone with lean combustion and AF ratio above 15 with
(barg)
calculated average CO mole fraction at the outlet of 69 ppm. Air flow rate (kg/s) 0.07 0.07 –
NOx model was implemented for the verification model with ther- Excess air (%) 37 37 –
mal, prompt NOx and reburn options. Fig. 5(e) and (f) show the gener- TIT (°C) 920–960 1069 16.2–11.4
ation of thermal and prompt NOx respectively. The formation of thermal CO (ppm) 50–60 69 20–15
NOx (ppm) 34–20 4 88.2–80
NOx is mainly dependent on the temperature and the availability of free
 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268 267

Fig. 7. Brake power output and combustion excess air.

extend down to the dilution zone since most of the fuel was converted used to produce an optimum design configuration. The simulations
into CO and CO2. Average mole fraction of the total NOx formation at the were first performed on different 3D geometries of chamber heights
outlet was below 5 ppm. and flame tube diameter using species transport combustion model.
The chamber and MGT were tested for extended period at 0.5 barg The non-premixed combustion model with laminar finite-rate and the
MGT pressure and it was noticed that TIT did not exceed 880 °C as illus- turbulent eddy-dissipation equations was further applied with radia-
trated in the MGT temperature profile in Fig. 6(a). Using low pressure tion P1 model to optimize the flame tube geometry. Chamber geometry
domestic LPG tank, the maximum operation pressure was about of 600 mm height and 152 mm diameter and 50 mm flame tube has
0.7 barg. Fig. 6(b) shows the experimental CO and NOx emission results provided low CO emissions of 102 ppm and chamber exhaust tempera-
during MGT start-up (0.05 barg) and up to maximum pressure of ture of 1218 °C. Optimum flame tube configuration included three
0.7 barg at different TIT. Combustion temperature presents the main zones: premix, combustion and dilution. Each zone consisted of 4
factor that governs CO production. However, since temperature was rows of holes of 6, 8 and 10 mm diameters respectively with 350 mm
not measured experimentally at the different combustions zones, TIT dead zone between the combustion and dilution zones.
was used as a direct indication of combustion intensity. CO emission The chamber was fabricated and tested experimentally with turbo-
was slightly above 100 ppm during MGT start-up due to the low excess charger based two-stage MGT. LPG fuel was used for the preliminary
air and combustion temperature with low NOx emissions below 5 ppm. test of the chamber. The dilution zone has proven to be effective in
However, after the start-up, compressor speed elevated rapidly provid- cooling the combustion product gasses below 900 °C without disturbing
ing higher air pressure and flow rate, thus, resulting in more intensive the combustion process with low CO emissions below 100 ppm. A Tef-
flame at the combustion zone and lower CO emission but with slightly lon brake dynamometer was connected to the power turbine and max-
elevated NOx emission up to 34 ppm. Table 3 shows a comparison be- imum brake power was about 0.55 kW.
tween the results obtained from the CFD model and experiment with
LPG fuel at similar operating conditions. The CFD simulation has provid-
ed good prediction of the flame distribution across the flame tube. How- Acknowledgement
ever, temperature and CO emission values were slightly over estimated
while lower values were predicted for NOx emission compared to the The authors would like to thank the Universiti Sains Malaysia (RUI
experimental results. grant number: 1001/PMEKANIK/814282) for the financial support of
this study.
6. MGT performance
References
A by-pass valve between the two stage turbines was used to release [1] G. Xiao, T. Yang, H. Liu, D. Ni, M.L. Ferrari, M. Li, Z. Luo, K. Cen, M. Ni, Recuperators for
some of the power turbine pressure to prevent shaft over speeding micro gas turbines: a review, Appl. Energy 197 (2017) 83–99.
when the brake is not engaged. The valve is closed during the brake [2] P.A. Pilavachi, Mini- and micro-gas turbines for combined heat and power, Appl.
Therm. Eng. 22 (2002) 2003–2014.
power testing only to prevent Teflon brake pad from overheating with [3] D. Chiaramonti, A.M. Rizzo, A. Spadi, M. Prussi, G. Riccio, F. Martelli, Exhaust emis-
additional vortex tube cooler to cool the pad. The MGT was operated sions from liquid fuel micro gas turbine fed with diesel oil, biodiesel and vegetable
with domestic low pressure LPG tank, hence, maximum MGT pressure oil, Appl. Energy 101 (2013) 349–356.
[4] A. Malmquist, Modeling and simulation of an externally fired micro-gas turbine for
was 0.7 barg. Fig. 7 shows the brake power and combustion excess air standalone polygeneration application, J. Eng. Gas Turbines Power 138 (2016) 1–15.
at different MGT pressures. Maximum brake power around 0.55 kW at [5] M. Jae, J. Ho, T. Seop, Program development and simulation of dynamic operation of
20,000 RPM, however, actual brake power is expected to be higher micro gas turbines, Appl. Therm. Eng. 108 (2016) 122–130.
[6] K.A. Al-attab, Z.A. Zainal, Performance of a biomass fueled two-stage micro gas tur-
due to leakage from the by-pass valve. The power turbine will be con- bine (MGT) system with hot air production heat recovery unit, Appl. Therm. Eng. 70
nected to a high speed alternator in the future development of the (2014) 61–70.
unit and by-pass valve will not be needed. [7] N.K. Singh, P. Tiwari, J. Srivastava, U. Pradesh, Simulation of microturbine generation
system for unbalance grid power supply, Int. J. Eng. Res. Technol. 2 (2013) 967–974.
[8] G. Ofualagba, The modeling and simulation of a microturbine generation system, Int.
7. Conclusion J. Sci. Eng. Res. 2 (2012) 1–7.
[9] M. Rabiu, M.B. Uday, A. Mohd, N. Ahmad, S. Rajoo, Microstructural evaluation of a
The design and development of MGT combustion chamber was per- slurry based Ni/YSZ thermal barrier coating for automotive turbocharger turbine ap-
plication, JMADE 109 (2016) 47–56.
formed, using CAD SOLID-WORKS and ANSYS-FLUENT 16.1 CFD simula- [10] A. Romagnoli, R. Martinez-botas, Heat transfer analysis in a turbocharger turbine: an
tion software. Various design parameters and combustion models were experimental and computational evaluation, Appl. Therm. Eng. 38 (2012) 58–77.
268 I.I. Enagi et al. / Fuel Processing Technology 166 (2017) 258–268

[11] L. Barelli, G. Bidini, F. Bonucci, Diagnosis of a turbocharging system of 1 MW internal [22] R. Gunadi, R. Camalia, CFD simulation of turbulent flows in proto X-3 bioenergy
combustion engine, Energy Convers. Manag. 68 (2013) 28–39. micro gas turbine combustor using STD k-ε and RNG k-ε model for green building
[12] A. Mehra, X. Zhang, A.A. Ayón, I.A. Waitz, M.A. Schmidt, C.M. Spadaccini, A six-wafer application, Int. J. Technol. 2 (2016) 204–211.
combustion system for a silicon micro gas turbine engine, IEEE J. [23] K. Zhang, A. Ghobadian, J.M. Nouri, Comparative study of non-premixed and partial-
Microelectromechanical Syst. 9 (2000) 517–527. ly-premixed combustion simulations in a realistic Tay model combustor, Appl.
[13] A. Mehra, A.A. Ay, I.A. Waitz, M.A. Schmidt, Microfabrication of high-temperature Therm. Eng. 110 (2017) 910–920.
silicon devices using wafer bonding and deep reactive ion etching, IEEE J. [24] H.S. Wah, Gas for Households-LPG for Cooking as First Step, Petroliam Nasional
Microelectromechanical Syst. 8 (1999) 152–160. Berhad, PETRONAS, 2013 http://www.ecreee.org/sites/default/files/documents/
[14] W.M. Yang, Microscale combustion research for application to micro news/day-1_1200_sookwah_lpg_for_cooking_final_0.pdf (accessed 01.04.07).
thermophotovoltaic systems, Energy Convers. Manag. 44 (2003) 2625–2634. [25] K.A. Al-attab, Z.A. Zainal, Syngas production and combustion characteristics in a bio-
[15] W.M. Yang, Combustion in micro-cylindrical combustors with and without a back- mass fixed bed gasifier with cyclone combustor, Appl. Therm. Eng. 133 (2017)
ward facing step, Appl. Therm. Eng. 22 (2002) 1777–1787. 714–721.
[16] H.L. Cao, J.L. Xu, Thermal performance of a micro-combustor for micro-gas turbine [26] D. Menghini, F.S. Marra, C. Allouis, F. Beretta, Effect of excess air on the optimization
system, Energy Convers. Manag. 48 (2007) 1569–1578. of heating appliances for biomass combustion, Exp. Thermal Fluid Sci. 32 (2008)
[17] C.P. Mark, A. Selwyn, Design and analysis of annular combustion chamber of a low 1371–1380.
bypass turbofan engine in a jet trainer aircraft, Propuls. Power Res. 5 (2016) 97–107. [27] K.A. Al-attab, Z.A. Zainal, Design and performance of a pressurized cyclone combus-
[18] M.M. Noor, A.P. Wandel, T. Yusaf, Design and development of mild combustion tor (PCC) for high and low heating value gas combustion, Appl. Energy 88 (2011)
burner, J. Mech. Eng. Sci. 5 (2013) 662–676. 1084–1095.
[19] M.M. Noor, A.P. Wandel, T. Yusaf, Detail guide for Cfd on the simulation of biogas [28] T. Furuhata, S. Amano, K. Yotoriyama, M. Arai, Development of can-type low NOx
combustion in bluff-body mild burner, Int. Conf. Mech. Eng. Res. (2013) 1–3. combustor for micro gas turbine (fundamental characteristics in a primary combus-
[20] G.C. Krieger, S. Paulo, F.L.S. Filho, R.C. De Souza, A swirler stabilized combustion tion zone with upward swirl), Fuel 86 (2007) 2463–2474.
chamber for a micro-gas turbine fuelled with natural gas, J. Braz. Soc. Mech. Sci. [29] B. Bazooyar, S.H. Hashemabadi, A. Shariati, NOX formation of biodiesel in utility
Eng. 34 (2012) 441–449. power plant boilers; part B. Comparison of NO between biodiesel and petrodiesel,
[21] A. Al-Halbouni, A. Giese, M. Flamme, K. Goerner, Applied modelling for bio- and lean Fuel 182 (2016) 323–332.
gas fired micro gas turbines, Prog. Comput. Fluid Dyn. 6 (2006) 391–405.