optimal heavy fuel direct injection analysis in a rotary ... · pdf fileoptimal heavy fuel...

1

American Institute of Aeronautics and Astronautics

Optimal Heavy Fuel Direct Injection analysis in a Rotary

Engine using a Computational Combustion Model

Asela A. Benthara W.1, Haibo Dong

2

Department of Mechanical and Materials Engineering

Wright State University, Dayton, OH 45435

Gregory R. Minkiewicz 3

1950 5th

Street

WPAFB, Ohio 45433

A numerical computational model was developed to replicate the single rotor

rotary engine with all three chambers. A Mesh was generated using GAMBIT 2.4.

Dynamic mesh motion was defined using a user defined motion profile inside the

ANSYS FLUENT. The combustion model used can represent both injection and

spark ignition models simultaneously. A parametric study was performed followed

by an optimum injection analysis to understand the effects of the injector and spark

plug when located at different locations.

Nomenclature

DISI = Direct Injection Spark Ignition

K = Turbulent kinetic energy

ε = Turbulent dissipation

CFD = Computation Fluid Dynamics

DPM = Discrete phase modeling

DOE = Design of Experiments

PISO = Pressure implicit with splitting of operators

bTDC = Before top dead center

TDC = Top dead center

aTDC = Before top dead center

CA = Crank angle

CO2 = Carbon Di Oxide

I. Introduction

he main objective of this computational study is to explore the optimum fuel injection for a 0.2 liter

direct injection spark ignition (DISI) Wankel rotary engine1 at 8000rpm using diesel fuel. Compared

to reciprocating type engines rotary engines have higher power to weight ratio. They are mechanically

simple, less vibrate and achieve better performance at high rpm2, 3 and 11

. Several port fuel injected rotary

engines can be found in the industry. Due to the inherent low fuel efficiency of rotary engine and increasing

gas prices, applications of the rotary engine in conventional automobiles is decreasing. The main goal of

this study is to improve the performance of the rotary engine in order to use it as an economical and

efficient power source for aero and automotive applications in the future.

Most recently achieved DISI technology is not a new idea. Engineers knew many advantages of

this technology such as controlled combustion, multi-fuel capability, improved fuel efficiency etc. But

early technology was not enough to achieve those designs. DISI technology required relatively higher

operating pressures because fuel is directly injected in to the combustion chamber. Injectors and sparks

should have high response times to achieve correct injection and spark timings. Current technological

advancements are sufficient to accomplish these goals. These capabilities are ideal for a rotary engine to

improve its efficiency. Several NASA publications related to the stratified charge DISI rotary engine

1 Graduate Student ,AIAA member, [email protected]

2 Associate Professor, AIAA Associate Fellow, [email protected]

3 Aerospace Engineer, Air Force Research Laboratory

T

2


computational and experimental research can be found in the reference reference4-6

. Some test results and

positive efforts for designing an experimental stratified charge rotary engine can be found in Curtiss-

Wright’s military rotary engine publication7. Compared to present publication similar approach for locating

the injector and spark can be seen in the NASA study but only few injection and spark locations were

considered for the analysis. In the present study an extensive optimization process were carried out

considering all the possible variations for the most important parameters to find the most suitable locations

to place an injector and spark. Six plumes of injection as opposed to single plume injection were deigned to

represent the spray from the injector more accurately. Also in this study diesel fuel, as opposed to gasoline,

is introduced into the rotary engine using DISI technology.

High-end technologies are required to test this type if DISI engines but it is very expensive to test

several engine designs at the same time in the real life. A good numerical simulation can make design

engineer’s life easier by saving a lot of money and time. Another goal of this project is to numerically

observe the fuel distributions at different injections and then combustions at different sparks to evaluate

possible power outputs. For this purpose an accurately designed 3D computational fluid dynamic model

will be used. ANSYS FLUENT computational fluid dynamic software and GAMBIT meshing software

will be used to serve these computational needs.

Design space for a rotary engine is large. In order to reduce total number of parameters, the most

important features for the combustion in the engine such as injection and spark are considered for the

optimization process. Commercially available 0.2L port fuel injected rotary engine model, sparks and

commercially available direct injectors are considered for the simulation. Experimental results were

obtained for the injectors and spark plugs in order to use in the computational model. Five different

parameters related to the injection and sparks were identified as important for the full factorial statistical

analysis. They are Injection location, Injection orientation, Injection timing, spark location and spark timing.

This full factorial design was followed by a statistical sensitivity analysis using JMP 800 commercial

software to determine how parameters affect the optimum injection and spark locations. The most sensitive

parameters and the less sensitive parameters were analyzed to find the optimum setup for single injection

rotary engine combustion. Contour plots of fuel consumption, generated CO2, equivalence ratio, average

temperatures and pressures were used to describe the results.

II. Methodology

A. Geometry Construction and Mesh Modeling

A commercially available rotary engine model was designed using Solidworks. Then this

simplified geometry was imported in to GAMBIT for mesh generation. Tetra/Hybrid element type was

used to create the mesh. Total number of elements was around 115600. Boundary conditions and fluid

quntinuumm types for each and chamber were assigned inside the GAMBIT.

Figure 1 Geometry modeling, mesh generation and FLUENT simulation

3


ANSYS FLUENT commercial software were used for the CFD simulation. Generated mesh was imported

into ANSYS FLUENT 12.1 software. Mesh motion and cold flow simulations were setup prior to running

any injections or combustion. Figure 1 explains the basic steps of the geometry, mesh modeling and CFD

simulation.

B. Simulation Setup

When setting up the dynamic mesh motion inside the FLUENT, a user defined mesh motion

profile was used. Numerical solver settings include unsteady pressure based solver with realizable k- ε

turbulence model. For the controlled combustion using DISI technology required both injection and spark

ignition models. Partially premixed combustion model supports both injection and spark models

simultaneously which were used during this simulation. Six plumes of injection8, 9

setup include solid cone

injection with Rossin-Rammler diametric distribution and droplet break up model. Other parameters related

to the rotary engine are shown in table 1.

Displacement (cm^3) ~200

Generating Radius (R) (mm) 71.5

Eccentricity(e) (mm) 11.6

Engine Speed (rpm) 8000rpm

Minimum Volume mm^3 25153.4

Maximum Volume mm^3 234145.0

Geometric Compression Ratio ~10

Table 1 Rotary Characteristics

C. Parametric Study and the Statistical analysis

Number of parameters affect the combustion was very large when testing a new engine design. By

following literature4-6, 8

and previous research related to the rotary engines helped to reduce most of the

parameters for the initial case setup. Five parameters were identified as the most important for the first

parametric study. For those five parameters, 2 levels for each and every parameter were selected as shown

in the table 2. A full factorial experimental design, consisted of 32 different simulations was run to identify

the effects of the spark and injection locations. Those five parameters and their two variations picked for

the case study were not the only constraints that affect a good combustion. Further analyses were conducted

with more levels to discover somewhat optimum setup for DISI rotary engine at 8000 rpm.

# Parameter Level 1 Level 2

X1 Spark position(from TDC) 0deg 10deg

X2 Spark start angle 0bTDC 10bTDC

X3 injector location(from TDC) -15deg -30deg

X4 injection orientation in to the flow With the flow

X5 start angle(CA deg) 180bTDC 90bTDC

Table 2 Five most important parameters and their variations

4


Injection parameters used for the general injection setup were matched from the solstice direct

injector as shown in table 3. Injector locations, orientations and timings were included in the parametric

studies as discussed above. Spark timings and locations were included as the other parameters for the

sensitivity study. 0.1 J spark energy was used to represent the spark from the spark plug. In FLUENT

spark was used to initialize the combustion. According to ANSYS, in the transient-partially premixed

solver, spark model was setup to solve chemical reactions.

General parameters related to the six plumes of injection

Temperature 300K

Flow rate per plume 0.002kg/s

Cone angle per plume 8deg

velocity magnitude 60m/s

Radius of a plume 0.000432m

Number of plumes( similar setup ) 6

Table 3 Injection parameters used in the simulation

For 5 parameters with 2 levels per each and every parameter, total of 25 = 32 simulations could be

designed. Those outputs then imported in to JMP800 statistical experimental design software for sensitivity

analysis. Outputs considered for the study included maximum average pressures and temperatures obtained

from ANSYS FLUENT post processing. For each and every simulation, diesel combustions and generated

CO2 was visualized to analyze the completeness of the combustion. JMP800 software was used to analyze

the data to get an idea about the sensitivity. DOE screening design for 5 parameters with 2 levels were

selected to construct the 32 cases. This analysis included full factorial design with full resolution by

considering all the variations of the picked parameters. Simple linear assumption was made when

following the JMP800 statistical model.

III. Results

A. Data analysis using computational outputs.

After running total of 32 full factorial simulations following outputs were obtained as shown in

table 4. Residual plots shown in figure 2 could be used to verify linear assumption made during the DOE

design. From figures 3 and 4it can be identified that parameter X4, X3 and X5 as the most sensitive factors

for the combustion, which were injector orientation, injector location and injector timing. According to

JMP800, spark location or spark timing was less sensitive compared to the injection. Injection was directly

affected the fuel distribution. This suggested that the best fuel distribution provided the best results from

the engine. This could be evaluated from further analysis using three injection parameters. Air fuel mixing

were described using particles and contours of equivalence ratio plots. Combustions were evaluated using

generated CO2 and left over fuel after the combustion plots.

5


Table 4 mass average maximum pressures and temperatures obtained for sensitivity studies

-15

6


Figure 2 Outputs from JMP800. Residual by predicted plot pressure (top left), temperature (bottom left), Half normal

plot pressure (top right), temperature (bottom right)

Figure 3 Sensitivity studies for mass average pressure from JMP 800

Results were again divided in to eight groups considering X4, X3 and X5 as shown in tables 5 and

6. Observations from the figure 4 explain how those 8 groups divided in to sub categories using average

pressure data. Several good combustions were identified and further analysis was done to identify the

important characteristics for those combustions.

7


Table 5 observations from the outputs

Figure 4 Pressure results from the sensitivity studies

Group Injector

location

injector

orientation

Injector

timing Cases Combs:

A -15 to 180 1,9,17,25 Normal

B -15 to 90 2,10,18,26 No

C -15 with 180 3,11,19,27 No

D -15 with 90 4,12,20,28 No

E -30 to 180 5,13,21,29 Good

F -30 to 90 6,14,22,30 Normal

G -30 with 180 7,15,23,31 Weak

H -30 with 90 8,16,24,32 No Table 6 Observations for 8 groups

Contour plots for good combustion were shown in figure 5. According to the figure relatively

good uniform air fuel mixture could be seen close to the mid pocket area. There were almost no fluid

particles left in the mixture which means all the injected fuel liquid particles were vaporized to make an

ideal air-fuel mixture for the combustion. More CO2 can be seen in the leading edge and in the middle of

the pocket area, which means all the fuel is consumed during the combustion process. By using moment

outputs from fluent, power outputs can also be calculated for each and every cycle. Cases in the group E

achieved the highest power output compared to the rest of the seven groups. Maximum power output

achieved in group E is around 30.5 hp at 8000rpm.

Figure 5 left -contours of equivalence ratio, middle- fuel particles left before the combustion,

right – generated CO2 after the combustion for good combustion case in Group E

Similar evaluations can be seen in figure 6 and 7 for the weak, normal and no combustion cases

based on the fuel distribution, CO2 generation and droplet particles left in the combustion chamber.

Particles left before the spark event is not shown in the table 8 because amount of fluid particles left is

negligible for those three groups. For the no combustion cases CO2 contour plots are not shown because

8


combustion is not enough to show any generated CO2. For the groups that had any combustion, figure 9

showed fuel consumptions versus generated CO2. Also figure 9 can be used to evaluate amount of diesel

fuel left after combustion to understand where to locate the injectors and spark plugs. Group C did not

have combustion due to fuel dislocation in the leading edge.

Contours of Phi Contours of CO2 Observations

Group A, F -Normal combustion

More of a uniform

air fuel distribution

can be seen in the

leading edge. But

not enough air to

combust all the fuel

concentrated in the

leading edge.

Group G- Weak Combustion

Concentrated fuel in

the far leading edge,

Incomplete

combustion.

Figure 6 Normal and weak combustion cases comparison using contours of PHI and CO2

Contours of Phi Particles left colored by Phi

Group B,D,H- No Combustion

Delayed injection

caused a weak fuel

air mixture. Due to

injection orientation

a stream of fuel

droplets can be seen

in the middle of the

pocket area.

Figure 7 No combustion cases observed using contours of PHI and particles left in the chamber

.

Combined Pressure plots and temperature plots were shown in figure 8 for groups E, A F and G.

This could clearly describe the difference between good combustions and weak combustions. Both plots

have same trend which helps to prove those observations done in table 6.

9


Figure 8 average pressure and temperature comparison

Group E-Good Group A-Normal

Group F-Normal Group G- Weak

Figure 9 fuel consumption and generated CO2for different combustions

B. Parameters and the levels of parameters evaluation.

It was important to re-evaluate the optimum parameters and their levels to understand the optimum

setup to locate a single injector in the combustion chamber. Parameters related to the injection came as the

most important from the sensitivity studies. Since this parametric study involved only two levels per

10


parameter, other possible levels should evaluate for the optimum setup. Possible levels for injection

parameters were described in the next section.

I. Injection orientation

Two orientations were considered as the first step in this parametric study. They were the injection towards

the flow and with the flow. Contours of velocity vectors during the injection event were shown in figure 10

to understand to physics behind the best orientation. In figure 10 black long arrows show the flow direction

according to the velocity vectors. Short red arrow in the chamber shows the injector orientation.

To the flow orientation With the flow orientation

Figure 10 Turbulence generation due to the orientation of the injection

When the injector was oriented opposed to the flow direction as shown in the arrow in figure 10 left picture,

larger area of turbulence generation could be seen compared to the right side figure which was the injection

oriented with the flow direction. This helped to generate a good air-fuel mixture during the short period of

time till the spark event (figure 5 and 6 groups E, A and F). When injector is oriented with the flow

direction, instead of mixing fuel with air, it tends to follow the air direction towards the leading edge of the

combustion chamber (figure 10). According to this results when injector is oriented in to the flow direction

will provide best air-fuel mixtures for the combustion.

II. Injection timing

More levels for injection timings were considered after evaluating the first set of parameters. Results

were shown in the table 7. From results when injection started at 160bTDC, horse power output seems

much higher compared to the rest of the levels. So 160bTDC injection could be used as the optimum timing

for the injection.

Start of injection Horse power output

bTDC Crank angle

200 160 29.49 180 180 30.94 160 200 33.41 120 220 31.54 100 240 31.92 90 260 29.90

Table 7 more levels for injection timing

11


III. Injection location

Injection location was the next important parameter according to the sensitivity studies. Figure 11 described

similar evaluation like injection timing to obtain the best injection location. 30 degrees below the TDC

injection location provided the best horse power output according the graph.

Figure 11 Injection location versus. horse power output

IV. Summary

As a new trend in the automobile manufacturing DISI technology could be used to control the

combustion and get higher gas mileage. This study explored how to optimize DISI technology to achieve

higher performance in the rotary engine. Unlike experimental engine tests CFD provided a very clear

picture in the engine combustion process.

When designing a new engine, number of parameters needed to be considered for the optimum

performance was large. By choosing few most important parameters design space can be reduced, but after

obtaining optimums for those picked small number of parameters it is important to verify results

considering all the possibilities of the other related parameters and their levels. After re-evaluating all the

levels 160bTDC injection start timing delivered the best power outputs with 30deg below injection location

and into the flow injection orientation. Combustion output is depending on the injection more than the

spark event. As long as air fuel mixture is homogeneous and located in the middle and leading side of the

pocket area, spark event closer to the TDC provided the maximum performance from the engine. This was

a good approach for optimizing a new engine design using CFD simulations to reduce number of

experimental test trials in the real life.

Acknowledgments

The authors would like to thank Brain Nicholson and Daniel Gillaugh, for providing the cad geometry

and injection specs for the simulation; Zongxian Liang for making the user defined motion profile and

Wright State University for the funding support.

29.60

29.80

30.00

30.20

30.40

30.60

30.80

31.00

31.20

10 15 20 25 30 35 40

ho

rse

po

we

r

location (deg)

12


References

1. Yamamoto, K. (1981). Rotary Engine. Tokyo, Japan: Sankaido Co. Ltd.

2. Ritsuharu, S., Shimizu, T., Tadokoro, N., & Junichi, F. (1992). Mazda 4-Rotor Rotary Engine for

the Le Mans 24-Hour Endurance Race. SAE technical Paper Series , 920309.

3. Shimizu, R., Tadokoro, T., Nakanishi, T., & Funamoto, J. (1992). Mazda 4-Rotor Rotary Engine

for Le Mans 24-Hour Endurance Race. Mazda Motor Corp , SAE Technical Paper Series, 920309.

4. Abraham, J., Bracco, F., & Epstein, P. (1993). 3-D Computations to improve Combustion ina

Stratified- Charge Rotary Engine Part4:Modified Geometries. SAE Technical Paper Series ,

930679.

5. Irion, C., & Mount, R. (1990-1992). Stratified Charge Rotary Engine Critical Technology

Enablement Volume1. New Jersey: NASA.

6. Raju, M., & Willis, E. (1991). Three-Dimensional Analysis and Modeling of a Wankel Engine.

SAE Technical Papers Series , 910701.

7. Jones, C. "Advanced Development of Rotary Stratified Charge 750 and 1500 HP Military Multi-

Fuel Engines at Curtiss-Wright", SAE Paper 840460, Feb. 1984.

8. Moda, S. U. (2010). Computational modeling and Analysis of Heavy Fuel Feasibility in Direct

injection Spark Ignition Engine. Dayton: Wright State University .

9. Moda, S. U., Dong, H., & Wan, H. (2011-Jan 4-7). Computational Investigation of Optiimal Fuel

Injection in Direct Injection Engine. AIAA Aerospace Science Meeting. Orlando: AIAA.

10. Richard, S. (1993). Introduction to Internal Combustion Engines Second Edition. Warrendale,

Pennsylvania: Society of Automotive Engineers, INc.

11. Benthara,A.A. (2010). Computational Investigation of Optimal Heavy Fuel Direct Injection Spark

Ignition in Rotary Engine. Dayton: Wright State University

optimal heavy fuel direct injection analysis in a rotary ... · pdf fileoptimal heavy fuel...

Documents