VALIDATION OF A 0D/1D COMPUTATIONAL CODE FOR

THE DESIGN OF SEVERAL KIND OF INTERNAL

COMBUSTION ENGINES

E. J. López and N. M. Nigro

Centro Internacional de Métodos Computacionales en Ingenierı́a (CIMEC)

Instituto de Desarrollo Tecnológico para la Industria Quı́mica (INTEC)
Universidad Nacional del Litoral - CONICET
Güemes 3450, 3000-Santa Fe, Argentina
ph: (54342)-4511 494/5, fx: (54342)-4511 169
[email protected], [email protected]

Abstract— A code for computational simu- models, the engine is represented as a network of pipes
lation of internal combustion engines is pre- (intake and exhaust manifolds) interconnected among
sented. One-dimensional gas dynamics equa- them with “devices” that simulate different parts of
tions are used for model the flow through the machine (valves, cylinders, pipe junctions, etc.).
pipes and manifolds, and the remaining com- One-dimensional CFD models are used for pipes and
ponents in the engine (cylinders, valves, etc.) thermodynamic (or zero-dimensional) models for the
are modeled by using thermodynamic or 0D above mentioned “devices”.
models. The numerical code developed is able For 0D models, most properties are averaged over
to simulate spark-ignition and compression- the total volume and no spatial information is avail-
ignition, two-stroke and four-stroke, multi- able. A survey of thermodynamic models for cylin-
cylinder and multi-valve engines, naturally as- ders are presented by Blumberg et al. (1979), Mattavi
pirated or turbo-charged, and different geome- et al. (1980), Heywood (1980), among others. These
tries of the combustion chamber. The code was models rely on some understanding of the physics in-
implemented in the scripting language Python, volved and try to capture the main features of the
which is a dynamic object-oriented program- processes. By including the description of the most
ming language that offers strong support for in- important aspects, the models have performed sur-
tegration with other languages and tools. The prisingly well and are ideally suited for parametric
numerical methods used in the discretization of studies. A zero-dimensional single-zone model is ca-
the equations and implementation details are pable of predicting engine performance and fuel econ-
presented. Several test cases are included in omy accurately with a high computational efficiency
order to show the performance of the code. (Krieger and Borman, 1966; Foster, 1985; Assanis and
Heywood, 1986). The major drawback of single-zone
Keywords— Internal combustion engine
models is their inability to simulate the wave propaga-
modeling, 0D/1D internal combustion engine
tion into pipes and manifolds that strongly influence
simulation, Python.
on volumetric efficiency. Also, these models are unable
to account for fuel spray evolution and the spatial vari-
I. INTRODUCTION ation in mixture composition and temperature, both
The modeling of reciprocating and rotary internal of which are essential in predicting harmful species
combustion (IC) engines is a multidisciplinary subject formed during the combustion process.
that involves thermodynamics, fluid mechanics, turbu- On the other hand, multi-dimensional models, such
lence, heat transfer, combustion, chemical reactions, as KIVA (Oran and Boris, 1981; Bracco, 1985; Ams-
mathematical analysis, and numerical methods. His- den et al., 1985; Amsden et al., 1987; Varnavas and
torically, different levels of approximation have been Assanis, 1996), resolve the cylinder space into fine
used to predict the performance of IC engines, from grids, thus providing a considerable amount of spa-
simple air standard cycles to complex 3D models in- tial information. However, multi-dimensional mod-
cluding turbulence, chemical reactions, spray dynam- els still employ phenomenological submodels describ-
ics, etc. IC engine simulation can be classified into ing fuel spray processes, and their simulation re-
four categories, namely zero-dimensional single zone, sults may vary with assumed initial or boundary
0D/1D single zone models, quasi-dimensional multi- conditions. Consequently, the accuracy of the re-
zone models and multidimensional models. In 0D/1D sults cannot always be guaranteed. Furthermore,
computational time and storage constraints still pre- coupled with the KIVA-3V (Amsden et al., 1993) code.
clude these models from routine use for design pur- The coupling was done applying a boundary condi-
poses. Currently an intermediate step between zero- tion for the 3D code based on the mass flow rate and
dimensional and multi-dimensional models has arisen, the pressure obtained from the 0D/1D code, for which
called quasi-dimensional. Multi-zone models (Hiroy- was adopted an absorbing boundary condition strat-
asu and Kadota, 1976; Hiroyasu and Kadota, 1983; egy to compute the flow state at the coupling interface
Kyriakides et al., 1986; Yoshizaki et al., 1993; Koure- (Bella et al., 2003). Generally, CFD-3D codes employ
menos et al., 1997; Rakopoulos and Hountalas, 1998; an implicit scheme of integration in time, therefore, the
Jung and Assanis, 2005) can be effectively used to sim- availability of a version of the 0D/1D simulator with
ulate new technology engine combustion systems, by implicit integration gives greater generality to the code
combining the advantages of zero-dimensional mod- as a generator of boundary conditions.
els and multi-dimensional models. These models are The new code was written using the language
able to provide the spatial information required to pre- Python (van Rossum, 1990-2007) because it offers
dict emission products with significantly less require- strong support for integration with other languages,
ment on computing resources than for multidimen- such as C++, Fortran, etc. Also, Python offers
sional models. the advantages of the object-oriented programming
In this work we start to use 0D/1D models in order and has a clear and simple syntax. The numeri-
to obtain a computational tool that can predict with cal code developed is able to simulate spark-ignition
sufficient precision the performance of an IC engine and compression-ignition, two-stroke and four-stroke,
at a relatively low computational cost, with the tar- multi-cylinder and multi-valve engines, naturally aspi-
get of arriving to more sophisticated quasi-dimensional rated or turbo-charged. Also, different geometries of
models in the future. Usually, the 0D/1D engine simu- the combustion chamber are available.
lators employ explicit schemes for integration in time. As future work we will use this software tool as the
Hence, the 0D models are generally formulated accord- simulation stage of an optimization package that al-
ing to an explicit scheme. Because most of these 0D low to setup the engine for performance improvement.
models are non-linear and also the need to reduce the Moreover, our interest would be to use this kind of
time step of the simulation for stability reasons when approximation in real time simulation software.
explicit schemes are applied, choosing an implicit inte- The paper is organized as follows. The next sec-
gration can provide more robustness to 0D/1D codes. tion presents the mathematical models used for each
This project began with the development of a sin- component. Next, the numerical methods applied and
gle cylinder four-stroke spark-ignition engine simula- some details about the implementation are included.
tor. The mathematical model was based on a thermo- Finally, numerical results for different kind of engines
dynamic model for the cylinder and a one-dimensional are shown and some conclusions are presented.
gas dynamics description of the intake and exhaust
systems (Nigro et al., 1999). Then, with the target fo- II. MATHEMATICAL MODEL
cused on the prediction of more real situations, mod- A. Pipe model
els for pipe junctions in multi-cylinder configurations
One-dimensional unsteady flow equations are used for
were added. In order to check the suitability and relia-
modeling pipes and manifolds. In order to include ef-
bility of this computational tool in industrial applica-
fects like variable cross-section, viscous friction, and
tions, this code was rewritten in the compiled language
wall heat transfer, some source terms are added to the
Fortran 90/95 and several test cases were solved val-
inviscid gas dynamic model represented by the system
idating the results with measurements (Alianak and
of Euler equations. The resultant system of equations
Nigro, 2003). In the same reference cite an interest-
can be written as (Heywood, 1988)
ing optimization work was done, where the coupling
between the 0D/1D code and other mechanical design
∂ρ ∂(ρu) 1 dF
tools for sizing camshafts are shown. This optimiza- + =− ρu
∂t ∂x F dx
tion task demanded an extensive usage of our IC en- 2
∂(ρu) ∂(ρu + p) 1 dF 2
gine simulator showing as a by-side product its inher- + =− ρu − ρG
ent robustness. In addition of the overall performance ∂t ∂x F dx (1)
prediction of IC engines, our goal is also to be able ∂(ρE) ∂[(ρE + p)u] 1 dF
+ =− u(ρE + p) +
to make 3D simulations of in-cylinder flows. Due to ∂t ∂x F dx
the impossibility to simulate an IC engine completely D
+ q̇π
with 3D models, it is common to use 0D/1D codes to F
simulate the whole engine except the component that
where ρ is the density; p is the pressure; u is the fluid
will be studied in detail. Thus, in this application,
velocity; F is the pipe cross-section area;
the 0D/1D engine simulator could be used as a gen-
erator of dynamic boundary conditions. In this line u|u| D
of work, the Fortran 90/95 version of the code was G=f π
2 F
is the specific friction force, with the friction coeffi- transfer coefficient hc with expressions like the follow-
cient given by f = 2τw /ρu2 , τw being the viscous shear ing
stress at the pipe wall and D the equivalent diameter
hc L
of the pipe; E is the total specific energy of the fluid; Nu = = CReα Prβ (3)
and q̇ is the heat transfer per unit mass of fluid per λ
unit time. The total specific energy is related to the where L is a characteristic length; λ is the gas thermal
internal energy per unit mass e and specific kinetic conductivity; and C, α and β are constants.
energy as Then, the heat transfer rate to the walls is
1
E = e + u2
2 Q˙ht = Ahc (T − Twall ) (4)
The equation of state used here corresponds to the
where T is the temperature of gas into the cylinder,
ideal gas assumption with particular gas constant
and Twall is the temperature of the cylinder wall.
Rgas .
B.3. Heat release model
B. Cylinder model
In order to modeling combustion we use several ap-
A single-zone model is used to model the cylinder. In proaches and mathematical models, which have the
this model the charge is assumed to be a homogeneous goal to describe the actual heat release via combus-
mixture of ideal gases at all times. The equations of tion as exactly as possible by means of the so-called
the model are the conservation of mass and the first substitute heat release rates.
law of thermodynamics For spark-ignition engines, the mass fraction of
burnt gases (xb ) is computed by using a Wiebe func-
dm X
= ṁj tion (Heywood, 1988)
dt j " 
(2) s+1 #
d mb θ − θig
xb = = 1 − exp −c (5)
X
(me) = −pV̇ + Q˙ch − Q˙ht + hj ṁj m ∆θ
dt j

In Eq. (5) mb represents the mass of burnt gases; θ is

where m is the mass contained in the cylinder; ṁj is
the crank shaft angle; ∆θ is the duration of combus-
the instantaneous mass flow rate through the j-th in-
tion; θig denotes the angle at which burning starts; c
let/outlet (for instance, the mass flow rate through in-
and s are parameters, where s is designated the shape
take and exhaust valves, fuel addition, leakages, etc.);
parameter and c accounts for combustion efficiency.
e is the specific internal energy of the mixture; V is
The heat release rate can be computed as
the cylinder volume; Q̇ch represents the heat release
due to combustion; Q̇ht is the heat transfer rate; and Q˙ch = Hc x˙b mf (6)
hj ṁj represents the enthalpy fluxes through the j-th
inlet/outlet. Hc being the calorific heat content of the fuel and mf
The model is closed specifying the geometry of the the total mass of fuel trapped into the cylinder.
combustion chamber, the heat release rate, the heat For compression-ignition engines, two substitute
transfer rate through the cylinder walls, and the mass heat release rates were implemented: the model pro-
flow rate of air and fuel. The sub-models used here posed by Watson et al. (1980) and the double Wiebe
are presented in the following sections. function (Miyamoto et al., 1985; Murayama et al.,
1982). In the model by Watson et al. the mass burning
B.1. Geometry of the combustion chamber
rate is expressed as
The geometric data necessary for the cylinder model
are the total surface area of the cylinder walls (A), the dxb ω
= [βc1 c2 τ c1 −1 (1 − τ c1 )c2 −1 +
volume of the cylinder and its time derivative. By the dt ∆θ (7)
0 0
time, it is possible to simulate conventional reciprocat- (1 − β)c01 c02 τ c1 −1 exp (−c01 τ c2 )]
ing engines, opposed-piston engines, and the MRCVC
(Motor Rotativo de Combustión a Volumen Constante; where τ = (θ − θig )/∆θ, and ω is the angular speed
Toth, (2004); Toth et al., (2000)). For all these cases, of the shaft. The coefficients proposed in the original
we implemented analytical formulae to compute the model are
geometric variables of the combustion chamber.
• c1 = 2 + 1.25 × 10−8 (N td )2.4 , where N is the
B.2. Heat transfer model engine speed, in rpm; and td is the ignition delay,
The instantaneous heat transfer rate that appears in ms.
in Eq. (2) is calculated applying Nusselt-Reynolds- • c2 = 5000.
Prandtl (Nu-Re-Pr) numbers correlations as, for ex-
ample, the one developed by Woschni (1967) or by • c01 = 14.2φ−0.644
ig , φig being the equivalence ratio
Annand (1963). All of them allows to compute a film at the time of ignition.
 • c02 = 0.79(c01 )0.25 . C. Valve model
−0.26 To calculate the flow rates through the intake and
• β = 1 − 0.926φ0.37
ig td .
exhaust valves, we use an analogy with the steady
The double Wiebe function is an extension of the flow through convergent nozzles proposed by Ben-
model used for spark-ignition engines, in order to de- son (1982). The model assumes the passage area
scribe the premixed and diffusive combustion periods through the valve as the nozzle throat (whose state
observed in diesel engines (Ramos, 1989). The mass is represented by the subscript T in the equations be-
fraction of burnt gases can be written as low), and the nozzle connecting the cylinder (subscript
C in the equations) and the end of the pipe (subscript
P in the equations). Depending on the direction of the
"  s +1 #
θ − θig p
xb = 1−xp exp −c − flow velocity with respect to the pipe end, the problem
∆θp
"  (8) could be an inlet (from cylinder to pipe) or an outlet
s +1 # (from pipe to cylinder). In addition, the flow at the
θ − θig di
xdi exp −c throat could be sonic or subsonic. The equations of
∆θdi
the model are presented below.
where xp is the mass fraction of fuel burnt in the pre- • Subsonic inlet:
mixed combustion period, xdi is the mass fraction of    
dp du
= RHS±


fuel burnt in the diffusive combustion period, ∆θp and ± ρP aP 2 P
∆θdi are, respectively, the duration of premixed and dt P dt P
diffusive combustion. ρT u T ψ = ρP u P
In diesel engines, the ignition delay time could be a2C = a2P + δu2P
calculated as the difference between the time at which  k (11)
combustion starts (tig ) and the time at which injection pC ρC
=
starts (tinj ). The time tig can be obtained from the pT ρT
following expression (Assanis, 1985): a2C = a2T + δu2T
Z tig pT = pP
dt
=1 (9)
tinj td where ψ = FT /FP , δ = (k−1)/2, k is the specific
heat ratio, a is the sound speed, and
which accounts for the pressure and temperature vari-  
D
ations resulting from compression. The ignition delay RHS± 2 =(k − 1) q̇π + ρuG ∓
time (td ) as a function of T and p was correlated for a F
(12)
variety of fuels with the expression ρua2 dF
ρaG −
F dx
td = Cp−n exp (Ta /T ),
In system (11), the first equation accounts for
the compatibility along the incoming Mach line
being C, n and Ta constants (Assanis, 1985). Also,
λ± ; the second equation is the mass conservation
the empirical formula developed by Hardenberg and
between T and P ; the third and fifth equations
Hase (1979) for predicting the ignition delay time in
represent the energy conservation between C and
direct-injection diesel engines was implemented in the
P , and C and T , respectively; the fourth equa-
code.
tion represents an isentropic evolution between
B.4. Scavenge model the cylinder and the nozzle throat; and the last
The scavenge process in two-stroke engines is modeled equation is the condition on the pressure at the
via a semi-empirical model proposed by Blair (1990). nozzle exit.
The mass of delivered air trapped into the cylinder • Subsonic outlet:
at exhaust closure (mat ) is computed as mat = ηS mt ,    
dp du
where mt is the total mass retained after the exhaust = RHS±


± ρP aP 2 P
port closure, and the scavenging efficiency (ηS ) can be dt P dt P
computed as ρT u T ψ = ρP u P
   
Dp 2 Dρ
ηS = 1 − exp (bσv + d) (10) − aP = (RHS1 )P
Dt P Dt P (13)
 k
b and d being constants experimentally determined, pP ρP
which depend on the type of scavenge. σv is the scav- =
pT ρT
enge ratio by volume, defined as the ratio between the
a2P + δu2P = a2T + δu2T
volume of air supplied during the scavenge period and
the cylinder volume. pT = pC
 with III. NUMERICAL IMPLEMENTATION
 
D The development of the code presented aims, in addi-
RHS1 = (k − 1) q̇π + ρuG (14)
F tion to its use as a tool for simulation of IC engines, to
be a generator of boundary conditions for CFD codes.
and D/Dt denoting the material derivative.
Due to the computational cost of the CFD simulations
From first equation to the last one in the sys-
on engines, its resolution is usually done using an im-
tem (13), they represent, respectively, the com-
plicit scheme of integration in time. Therefore, it is
patibility along the incoming Mach line, the
necessary that the 0D/1D engine simulator could work
mass conservation between T and P , the com-
with an implicit scheme avoiding a severe restriction
patibility along the incoming path line λ0 , the
on the time step.
isentropic evolution between P and T , the en-
The use of an implicit scheme for the discretization
ergy conservation between P and T , and the con-
of equations that model the several devices in the code
dition on the pressure at the nozzle exit.
do not represent a problem for its practical implemen-
• Sonic inlet: in this case, the system of equations tation, with the exception of models of valve and pipe
is the same as Eq. (11) with the last equation junction. As presented above, the systems of equations
replaced by the condition aT = uT . that model valves and pipe junctions change with the
flow regime and the direction of its velocity.
• Sonic outlet: in this case, the system of equations In the valve model, to determine whether the prob-
is the same as Eq. (13) with the last equation lem is an inlet or an outlet, we compare the cylinder
replaced by aT = uT . pressure with the stagnation pressure (p0P ) at the pipe
D. Pipe junction model end, which is given by (White, 1983)
The pipe junction model applied was proposed by Cor- "  2 #k/(k−1)
beran (1992). If the junction is composed by r incom- uP
p0P = pP 1 + δ (15)
ing pipes and s outgoing pipes, the model is expressed aP
as
• Mass conservation The valve is considered open when the passage area
is larger than a prefixed tolerance A > 0. Then, if
N
X pC > p0P , the flow is established from the valve to the
ṁj = 0, with ṁj = ρj Fj uj nj
pipe, otherwise it is considered an outgoing flow from
j=1
the pipe. To take into account the transition between
where N = r + s is the total number of pipes at subsonic and sonic regime flow, we use the following
the junction, Fj is the cross-section area of the convex combination
j-th pipe and nj its exterior normal.
R = (1 − χ)Rsonic + χRsubsonic (16)
• Energy conservation
N where Rsubsonic and Rsonic represent the systems of
X ṁj
h˙j = 0, with h˙j = (a2 + δu2j ) equations that model the subsonic and sonic cases, re-
j=1
k−1 j spectively; and
  
• Compatibility equation along incoming Mach 1 aT − uT
χ= 1 + tanh (17)
lines λ±
j 2 αaT
   
dp du α being a coefficient that adjusts the transition of χ
= RHS±


±ρj aj 2 j , j = 1, · · · , N
dt j dt j between 0 and 1. In valve model, the compatibil-
ity equations along the characteristic curves are used
• Compatibility equation along incoming path to complete the system of equations. These must be
lines λ0j solved according to an explicit scheme, implying that
    the states at the pipe end, the valves, and the cylin-
Dp 2 Dρ
− aj = (RHS1 )j , j ∈ r der, do not depend on the state at interior points in the
Dt j Dt j pipe at time t + ∆t. Thus, the system of equations for
valve+cylinder is decoupled from the remaining equa-
• Equality of pressure at all branches in the junc-
tions. However, the resolution is done in a coupled
tion
way with the goal of being able to implement other
pi = pj , ∀i 6= j
valve models of full implicit type in the code. When
• Equality of enthalpy at all outgoing branches in the valve is either opening or closing, it is important to
the junction determine the direction of the flow due to the change
in the system of equations that models each situation,
a2i + δu2i = a2j + δu2j , ∀i, j ∈ s, i 6= j which directly influences on the global convergence.
To predict the flow direction it is assumed an isentropic (MPI-Forum, 1994-2008). This feature can be useful
flow through a convergent nozzle between the cylinder to include the engine simulator in a parametric opti-
and the corresponding end of the pipe. The flow is mization strategy or to couple with other codes that
established between the states U0 and Ue , which are run in parallel.
assumed constant and are identified with either the Regarding the computational implementation, a
state in the cylinder (UC ) or the state at the end of class was defined for each device with methods that
the pipe (UP ), depending on the relationship between perform the computation of sub-models and the con-
the pressure in the cylinder and the stagnation pres- tributions to both global residue and global jacobian
sure p0P . If pC ≥ p0P , then it is adopted U0 = UC matrix. For instance, the class Cylinder contains the
and Ue = UP , the direction of the flow being from the methods combustion, geometry, heat transfer, etc.
cylinder to the pipe. When pC < p0P , then U0 = UP All of these classes derive from the classes that define
and Ue = UC , yielding to a flow from the pipe to the an specific set of equations for each device. For ex-
cylinder. The critical pressure ample, the class Flow involve the equations of mass,
 k/(k−1) energy and linear momentum conservation. This al-
2 lows to add new features to the code as, for example,
pcrit = p0 (18)
k+1 transport of chemical species with very few changes.
determines if the nozzle throat is choked or not. If Currently, only the class Flow is implemented in the
pe > pcrit , the predicted state at valve (UT ), i.e. the code. The constructor of each class defines attributes
nozzle throat, is taken as for the object, some of them are required data and the
remaining ones are optional (defined through default
pT = pe values).

pT
1/k The nonlinear global system of equations is solved
ρ T = ρ0 via the Newton-Raphson method. The linear sys-
p0 (19)
#)1/2 tem arising at each nonlinear iteration can be solved
( "  k−1 either using functions contained into the package
k pT p0 k
uT = −1 NumPy (Oliphant, 2006), or with the package petsc4py
δ ρT pT
(Dalcı́n, 2005-2007). petsc4py is a set of Python bind-
If pe ≤ pcrit , then it is assumed ings for PETSc, the Portable, Extensible Toolkit for
Scientific Computation (Balay et al., 1995-2007). Cur-
pT = pcrit
rently, we are working to improve the efficiency of the
 1/(k−1)
2 code using Cython (Behnel et al., 2007-2009).
ρ T = ρ0 The code also allows to define the parameters of
k+1 (20)
 1/2 calibration and the operational variables as functions
2k p0 of the engine speed, the cycles, and the time.
uT =
k + 1 ρ0
IV. RESULTS
To solve the system of equations (1), the Fi-
In the next sections we will present the results ob-
nite Element Method stabilized with the Streamline
tained from the application of the code to the simu-
Upwind/Petrov-Galerkin (SUPG) technique (Hughes
lation of some IC engines. The cases were selected to
and Mallet, 1986; Tezduyar and Hughes, 1983) was
show the spectrum of IC engines that can be simu-
used. Time derivatives were discretized applying the
lated.
trapezoidal finite difference scheme.
The code was implemented in the scripting language A. Four-stroke spark-ignition engine test
Python (van Rossum, 1990-2007). Python is a dy- The first example of application is a 8.4 l. V10 four-
namic object-oriented programming language that can stroke spark-ignition engine. Tables 1 and 2 contain
be used for many kinds of software development. It the main data of the engine.
offers strong support for integration with other lan-
guages and tools, and comes with extensive standard
libraries. Object-oriented programming allows to de- Table 1: Main cylinder data of the V10 engine.
velop the code in an organized manner, and the possi- Cylinder
bility of integration with other languages make it suit- Bore 103.0 mm.
able for solving the coupling between the 0D/1D sim- Stroke 100.6 mm.
ulator and CFD codes. Another feature that Python Connecting rod length 158.55 mm.
offers is the possibility of writing higher-level parts Compression ratio 9.6:1
of large-scale scientific applications and driving sim-
ulations in parallel architectures like clusters of PCs. The equivalence ratio, the coefficients of heat trans-
This can be done by using, for instance, the pack- fer through the cylinder walls, and the crank angle
age mpi4py (Dalcı́n et al., 2005) which provides bind- where the combustion starts, are functions of the en-
ings of the Message Passing Interface (MPI) standard gine speed. These functions were obtained from the
 0.5

Table 2: Valve data of the V10 engine. calculated

average mass flow rate of air [kg/sec]

0.45
experimental
Intake valve Exhaust valve 0.4

Diameter 52.8 mm. 40.89 mm. 0.35

Max. lift 14.4 mm. 13.59 mm. 0.3

VO 41◦ BTDC 129◦ BBDC 0.25
VC 123◦ ABDC 41◦ ATDC 0.2

0.15

0.1
experimental data available. The engine was tested at
speeds ranging from 1600 rpm to 6000 rpm. 0.05
1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Figures 1 to 3 show the computed indicated power, engine speed [rpm]

the torque and the average mass flow rate of air versus
the real curves obtained experimentally. Figure 3: Average mass flow rate of air at intake port
as a function of rpm.
800

calculated
700 experimental valve lift was defined in such a way to make the pas-
600 sage area computed by the code the same as the in-
indicated power [hp]

stantaneous passage area of the port. The complete

500
set of data can be found in the literature by Blair
400 (1990). Figure 5 shows the pressure in the cylinder,
300 cylinder

200

100
1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 transfer exhaust
engine speed [rpm] transfer port port exhaust
duct tank

Figure 1: Indicated power as a function of rpm.

intake
850 port
crankcase
calculated
800 fitted curve
experimental Figure 4: Computational model of the QUB 400 en-
indicated torque [Nm]

750 gine.
700
the crankcase, the transfer duct, and the intake and
650 exhaust ports as a function of the crank angle during
600 a cycle. Mass flows through transfer and exhaust ports
are shown in Fig. 6, where positives values represent
550
incoming flow to cylinder.
500
1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 TDC IPC EPO TPO BDC TPC EPC IPO TDC
engine speed [rpm] 200000
cylinder
180000 crankcase
Figure 2: Torque as a function of rpm. exhaust port
transfer port
160000
intake port
pressure [Pa]

140000
B. Two-stroke spark-ignition engine test
This test case was taken from the literature (Blair, 120000

1990) and consists in a two-stroke spark-ignition 100000

single-cylinder research engine denominated QUB 400.
80000
The engine speed is 3000 rpm at full throttle. To
model the crankcase compression we use a cylinder 60000
0 50 100 150 200 250 300 350
without combustion, then the computational model crank angle [deg]
is composed by two cylinders as can be observed in
Fig. 4. The engine has six transfer ducts between the Figure 5: Variation of pressure during a cycle.
cylinder and the crankcase, which were modeled as a
unique pipe with the same total cross-sectional area Table 3 shows some performance characteristics:
(see Fig. 4). Ports were assumed as poppet valves power, indicated mean pressure (IMEP), scavenge ef-
with appropriate discharge coefficients, and where the ficiency (ηS ), peak cylinder pressure (Pmax ) and the
 EPO TPO BDC TPC EPC
0.25 – Length: 150 mm.
transfer port
0.2 – Wall temperature: 292 K.
exhaust port
0.15
• Exhaust pipe data:
mass flow [kg/s]

0.1
– Diameter: 38 mm.
0.05
– Length: 400 mm.
0
– Wall temperature: 743 K.
−0.05

−0.1 • Intake valve data:

−0.15
0 50 100 150 200 250 300 350 – Diameter: 40 mm.
crank angle [deg]
– Maximum lift: 8.845 mm.
Figure 6: Mass flow rates through transfer and exhaust – IVO: 15◦ BTDC.
ports. – IVC: 50◦ ABDC.
• Exhaust valve data:
crank angle at which it occurs. In order to compare,
results reported by Blair (1990) and experimental data – Diameter: 40 mm.
are included. – Maximum lift: 8.803 mm.
– EVO: 65◦ BBDC.
Table 3: Performance characteristics of the QUB 400
– EVC: 15◦ ATDC.
engine.
Calculated Blair (1990) Exp. Table 4 contains the operational data of the three cases
Power [kW] 12.37 12.37 12.40 solved. In that table, pi and Ti are the intake manifold
IMEP [bar] 6.83 6.81 6.80 mean pressure and temperature, respectively; pe is the
ηS 0.77 0.84 - exhaust manifold mean pressure; and minj is the mass
Pmax [bar] 37.0 36.2 36.9 of fuel injected per cycle.
Angle [deg] 15.9 16.2 -
Table 4: Operational variables of the KamAZ-7405
These results were in a very good agreement with diesel engine.
those presented by Blair (1990) coming from his nu- Case N pi Ti pe minj
merical simulations and also with his experiments. [rpm] [bar] [K] [bar] g/cycle
C. Four-stroke diesel engine test 1 2200 1.97 390 1.51 0.0778
2 1400 1.52 354 1.36 0.0838
The test case selected to show the application
3 1000 1.28 338 1.20 0.0818
of the code to a diesel engine was the truck
diesel KamAZ-7405, taken from the web-page
http://www.diesel-rk.bmstu.ru/. The main en- The combustion was modeled using two Wiebe func-
gine data are the following: tions, and the ignition delay time was calculated with
the correlation proposed by Hardenberg and Hase.
• Number of cylinders: 8. The calibration of the code was done at 2200 rpm,
defining the parameters for the two Wiebe functions,
• Cylinder data: the coefficients of the heat transfer model, and the du-
– Bore: 120 mm. ration of the combustion. The results are presented
in Tab. 5 together with the relevant experimental re-
– Stroke: 120 mm. sults, where SFC is the specific fuel consumption and
– Connecting rod length: 225 mm. ṁa is the mean mass flow rate of air through the in-
– Compression ratio: 16:1. take system. There is in general a good agreement
with experimental data, being the error within typical
– Injection starting angle: 14◦ BTDC.
error margins for the type of code employed.
– Combustion starting angle: 9◦ BTDC.
V. CONCLUSIONS
– Duration of combustion: 69◦ .
A code for computational simulation of internal com-
– Wall temperature: 459 K. bustion engines was presented. The mathematical
• Intake pipe data: models and the numerical methods used were de-
scribed briefly. Some models and the solution strate-
– Diameter: 44 mm - 52 mm. gies were reformulated to use an implicit scheme of
 SAE International Congress and Exposition, De-
Table 5: Performance characteristics of the KamAZ-
troit, Mich., USA, SAE Paper N◦ 872072 (1987).
7405 diesel engine.
Power IMEP SFC ṁa Amsden, A.A., Ramshaw, J.D., O’Rourke, P.J. and
[kW] [bar] [g/kW h] [kg/s] Dukowicz, J.K., A Computer Program for two and
Calc. 193.4 11.69 212.4 0.326 three Dimensional Fluid Flows with Chemical Re-
1
Exp. 193.6 11.70 212.4 0.346 action and Fuel Spray, Los Alamos Labs., LS 12503
Calc. 139.5 12.08 201.8 0.174 MS (1993).
2
Exp. 138.8 12.03 202.8 0.182
Calc. 91.6 10.95 214.4 0.107 Annand, W., “Heat transfer in the cylinders of recipro-
3 cating internal combustion engines,” Proc. Instn.
Exp. 92.5 11.05 212.2 0.112
Mech. Engrs., 177, 973-980 (1963).

integration in time. The code was written in the lan- Assanis, D.N., A computer simulation of the tur-
guage Python in order to take advantage of the object- bocharged turbocompounded diesel engine system
oriented programming and the possibility of integra- for studies of low heat rejection engine perfor-
tion with other languages. Some test cases were solved, mance, Ph. D. Thesis, Massachusetts Institute of
including spark- and compression-ignition, two- and Technology (1985).
four-stroke IC engines. After code calibration, results
were in good agreement with experimental data. Assanis, D.N. and Heywood, J.B., “Development and
As future works we propose to use the code pre- use of a computer simulation of the turbocom-
sented as a generator of boundary conditions for CFD- pounded diesel system for engine performance and
3D codes through an appropriate strategy of coupling. component heat transfer studies,” SAE Interna-
In addition, we intend to incorporate the 0D/1D sim- tional Congress and Exposition, Detroit, Mich.,
ulator within a code of parametric optimization for IC USA, SAE Paper N◦ 860329 (1986).
engines.
Balay, S., Buschelman, K., Gropp, W.D., Kaushik,
ACKNOWLEDGMENTS D., Knepley, M.G., McInnes, L.C., Smith, B.F.
This work has received financial support from Con- and Zhang, H., PETSc: Portable, extensi-
sejo Nacional de Investigaciones Cientı́ficas y Técnicas ble toolkit for scientific computation (1995-2007),
(CONICET, Argentina, grants PIP 5271/05), Univer- http://www.mcs.anl.gov/petsc/.
sidad Nacional del Litoral (UNL, Argentina, grant
CAI+D 2005-10-64) and Agencia Nacional de Pro- Behnel, S., Bradshaw, R. and Sverre Seljebotn, D.,
moción Cientı́fica y Tecnológica (ANPCyT, Argentina, Cython: C-Extensions for Python (2007-2009),
grants PICT 12-14573/2003, PME 209/2003). Au- http://www.cython.org/.
thors made extensive use of freely distributed software
Bella, G., Rotondi, R., Defranco, G. and Nigro, N.,
as GNU/Linux OS, Python, G95 compiler, Octave,
“Code coupling tho enhance CFD analysis of IC
among many others. Special thanks are due to Lisan-
engines,” Mecánica Computacional, Vol. XXII, 26-
dro Dalcı́n for his support with the Python language,
40 (2003).
and also to Mario Storti for his contribution into the
discussion of the ideas. Benson, R., The Thermodynamics and Gas Dynamics
REFERENCES of Internal Combustion Engines, Vol. I, Clarendon
Press, Oxford (1982).
Alianak, J. and Nigro, N., “Intake and exhaust sys-
tem optimization of internal combustion engines,” Blair, G.P., The Basic Design of Two-Stroke Engines,
Technical Report, RT-ID-012, Departamento de In- Society of Automotive Engineers, Inc. (1990).
genierı́a, Facultad de Ciencias Exactas, Ingenierı́a
Blumberg, P.N., Lavoie, G.A. and Tabaczynsli, R.J.,
y Agrimensura, Universidad Nacional de Rosario
“Phenomenological models for reciprocating inter-
(2003).
nal combustion engines,” Prog. in Energy and
Amsden, A.A., Ramshaw, J.D., O’Rourke, P.J. and Comb. Sci., 23 (1979).
Dukowicz, J.K., KIVA: A Computer Program for
two and three Dimensional Fluid Flows with Chem- Bracco, F.V., “Modeling of engine sprays,” SAE Inter-
ical Reaction and Fuel Spray, Los Alamos Labs., LS national Congress and Exposition, Detroit, Mich.,
10245 MS (1985). USA, SAE Paper N◦ 850394 (1985).

Amsden, A.A., Butler, T.D. and O’Rourke, P.J., Corberan, J.M., “A new constant pressure model for
“KIVA-II computer program for transient multidi- N-branch junctions,” Proc. Inst. Mech. Eng., D
mensional chemically reactive flows with sprays,” J. automob. eng., 206, 117-123 (1992).
Dalcı́n, L., Paz, R. and Storti, M., “MPI for python,” and Exposition, Detroit, Mich., USA, SAE Paper
Journal of Parallel and Distributed Computing, N◦ 970635 (1997).
65(9), 1108-1115 (2005).
Krieger, R.B. and Borman, G.L., “The computation
Dalcı́n, L.D., PETSc for Python: Python of apparent heat release from internal combustion
bindings for PETSc libraries (2005-2007), engines,” ASME paper 66-WA/DGP-4 (1966).
http://code.google.com/p/petsc4py/.
Kyriakides, S.C., Dent, J.C. and Mehta, P.S., “Phe-
Foster, D.E., “An overview of zero-dimensional ther- nomenological diesel combustion model includ-
modynamic models for IC engine data analysis,” ing smoke and NO emission,” SAE International
SAE International Congress and Exposition, De- Congress and Exposition, Detroit, Mich., USA,
troit, Mich., USA, SAE Paper N◦ 852070 (1982). SAE Paper N◦ 860330 (1986).

Hardenberg, H.O. and Hase, F.W., “An empirical for- Mattavi, J.N., Groff, E.G., Lienesch, Matekunas, F.A.
mula for computing the pressure rise delay of a and Noyes, R.N., “Engine improvements through
fuel from its cetane number and from the rele- combustion modelling,” in Combustion Modeling
vant parameters of direct-injection diesel engines,” in Reciprocating Engines, Mattavi and Amann,
SAE International Congress and Exposition, De- Plenum Pres (1980).
troit, Mich., USA, SAE Paper N◦ 790493 (1979).
Message Passing Interface Forum,
Heywood, J.B., “Engine combustion modeling - An http://www.mpi-forum.org/ (1994-2008).
overview,” in Combustion Modeling in Reciprocat- Miyamoto, N., Chikahisa, T., Murayama, T. and
ing Engines, Mattavi and Amann, Plenum Pres Sawyer, R., “Description and analysis of diesel en-
(1980). gine rate of combustion and performance using
Heywood, J., Internal Combustion Engine Fundamen- Wiebe’s functions,” SAE International Congress
tals, McGraw-Hill, New York (1988). and Exposition, Detroit, Mich., USA, SAE Paper
N◦ 850107 (1985).
Hiroyasu, H. and Kadota, T., “Models for combus-
tion and formation of nitric oxide and soot in di- Murayama, T., Miyamoto, N., Yamada, T.,
rect injection diesel engines,” SAE International Kawashima, J.-I. and Itow, K., “A method to im-
Congress and Exposition, Detroit, Mich., USA, prove the solubility and combustion characteristics
SAE Paper N◦ 760129 (1976). of alcohol-diesel fuel blends,” SAE International
Congress and Exposition, Detroit, Mich., USA,
Hiroyasu, H., Kadota, T. and Arai, M., “Development SAE Paper N◦ 821113 (1982).
and use of a spray combustion modeling to pre-
dict diesel engine efficiency and pollutant emissions Nigro, N., Storti, M. and Ambroggi, L., “Modeliza-
(Part 1 combustion modeling),” Bull. Jap. Soc. ción numérica de un motor de combustión interna
Mech. Engrs., 26(214), 569-575 (1983). monocilı́ndrico,” Revista Internacional de Métodos
Numéricos para cálculo y diseño en Ingenierı́a,
Hiroyasu, H., Kadota, T. and Arai, M., “Development 15(1), 21-54 (1999).
and use of a spray combustion modeling to pre-
Oliphant, T.E., Guide to NumPy (2006),
dict diesel engine efficiency and pollutant emissions
http://numpy.scipy.org/.
(Part 2 computational procedure and parametric
study),” Bull. Jap. Soc. Mech. Engrs., 26(214), Oran, E.S. and Boris, J.P., “Detailed modeling of com-
576-583 (1983). bustion systems,” Prog. Energy Combust. Sci., 7,
1-72 (1981).
Hughes, T. and Mallet, M., “A new finite element
method for CFD: the generalized streamline oper- Rakopoulos, C.D. and Hountalas, D.T., “Develop-
ator for multidimensional advection-diffusion sys- ment and validation of a 3-D multi-zone combus-
tems,” Comp. Meth. Applied Mech. Eng., 58, tion model for the prediction of DI diesel engines
305-328 (1986). performance and pollutants emissions,” SAE Inter-
national Congress and Exposition, Detroit, Mich.,
Jung, D. and Assanis, D.N., “Modeling of direct injec-
USA, SAE Paper N◦ 981021 (1998).
tion diesel engine emissions for a quasi-dimensional
multi-zone spray model,” Int. J. Automot. Tech., Ramos, J.I., Internal Combustion Engine Modeling,
5(3), 165-172 (2005). Hemisphere Publishing Corporation (1989).
Kouremenos, D.A., Rakopoulos, C.D. and Hounta- Tezduyar, T. and Hughes, T., “Finite element formula-
las, D.T., “Multi-zone combustion modeling for the tions for convection dominated flows with particu-
prediction of pollutants emissions and performance lar emphasis on the compressible Euler equations,”
of DI diesel engines,” SAE International Congress AIAA, 83-0125 (1983).
Toth, J., Di Nezio, J., Staniscia, C. and López,
E., “Ventajas mecánicas y termodinámicas de un
nuevo motor rotativo,” 9 ◦ Congreso Chileno de In-
genierı́a Mecánica, Valparaı́so, Chile (2000).

Toth, J.A., Motor Rotativo de Combustión a Vo-

lumen Constante (MRCVC), Patent Res. N◦
AR004806B1, Rec. N◦ P 19960105411 (2004).

van Rossum, G., Python programming language (1990-

2007), http://www.python.org/.

Varnavas, C. and Assanis, D.N., “A high temperature

and high pressure evaporation model for the KIVA-
3 code,” SAE International Congress and Exposi-
tion, Detroit, Mich., USA, SAE Paper N◦ 960629
(1996).

Watson, N., Pilley, A.D. and Marzouk, M., “A com-

bustion correlation for diesel engine simulation,”
SAE International Congress and Exposition, De-
troit, Mich., USA, SAE Paper N◦ 800029 (1980).

White, F.M., Mecánica de fluidos, McGraw-Hill de

México (1983).

Woschni, G., “A universally applicable equation for

the instantaneous heat transfer coefficient in the
internal combustion engine,” SAE International
Congress and Exposition, Detroit, Mich., USA,
SAE Paper N◦ 670931 (1967).

Yoshizaki, T., Nishida, K. and Hiroyasu, H., “Ap-

proach to low NOx and smoke emission engines
by using phenomenological simulation,” SAE Inter-
national Congress and Exposition, Detroit, Mich.,
USA, SAE Paper N◦ 930612 (1993).