Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Mean Value Engine Modeling for a Diesel Engine 2011-01-1294

Published
with GT-Power 1D Detail Model 04/12/2011

Hai Wu and Xinlei Wang

Univ. of Illinois at Urbana-Champaign

Richard Winsor and Kirby Baumgard

John Deere Power Systems

Copyright © 2011 SAE International

doi:10.4271/2011-01-1294

spark ignition (SI) engines, while the cruise control was

ABSTRACT implemented and power density was improved. To balance
Mean value engine model (MVEM) is the basis of control the turbine at both low speed and high speed conditions,
design for advanced internal combustion engines. The engine wastegate or variable geometry turbocharger (VGT) control
performance transient process usually takes a few cycles. The became necessary. Exhaust gas recirculation (EGR)
MVEM provides an adequate accurate description of the technology and after-treatment systems were the main
engine dynamics with reasonable approximation by ignoring methods to reduce NOx emission from inside and outside the
the heat loss and sub-cycle events. MVEM is very important cylinders. Besides, the overlap vale, internal exhaust gas
for engine system control development, especially when the recirculation (EGR), high pressure fuel injection, and
modern engine becomes more and more complicated when multiple fuel injection technology may be necessary for the
equipped with throttle, turbocharger and after-treatment promising homogeneous charged compression ignition
systems. Usually the MVEM is developed based on data from (HCCI) engines. Furthermore, combustion model switching,
engine tests, which is a costly and time consuming process. In multistage turbochargers have been discussed more and more
this paper, the air path MVEM modeling method based on the for high performance in recent articles [1, 2, 3]. The
1D detail model is discussed for a turbocharged diesel engine. implementation of these new technologies depends on the
Simulation is applied to demonstrate the effectiveness of this control system to cooperate with the sub-systems, like the air
new method. This approach could be used to get the MVEM system, fuel system, combustion in chambers, and exhaust
for control design even before the prototype engine is system.
available. It reduces the cost, risk and labor compared with
the test data based approach. This MVEM model can be built Modern control theory depends on the system model much
in modules and the parameters can be validated for a specific more than classic control, like proportional-integrative-
engine. These advantages make it applicable to a wide range derivative (PID) control and frequency domain regulation.
of engines. While modern control theory provides more advanced
properties including robustness, optimization, multivariable
control, learning ability, and adaptivity. Nowadays, engine
INTRODUCTION modeling technology plays an important role in the engine
The ever increasing EPA regulations on vehicle emissions design and brings more challenges to control engineers.
and the fuel economy demands from markets promotes the Different control oriented modeling methods have been used
innovation for advanced engine technology. In past decades for different purposes in the implementation of advanced
electronic control has made a big contribution to the engine technology, such as a simple input-output block model
development of some new technologies. For example, to and the MVEM which consists of detail sub-systems. In
improve the engine performance and drivability, the addition, the crank angle based combustion model and fluid
electronic throttle control and turbocharger were applied in
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

dynamics based air charge model are needed for in-cylinder disadvantage of empirical and analytical models were
control. discussed in [7]. Then a mean value model of the gas
exchange was developed for use in power train applications.
In the modern diesel engine, the turbocharger is widely used
to increase power density. The turbocharger increases the In engine performance control, the most interested variables
complexity for the engine to deliver the expected torque to are engine and turbocharger torque, speed, fluid dynamics at
the crankshaft, while satisfying demands for drivability and throttle and manifold temperature, pressure, and mass flow
fuel economy, subject to emission constraints. However, an rate. The fast sub-cycle events like combustion and valve
improper choice of controller and parameters can lead to an timing may be ignored according to specific application.
undesirable torque response. A large amount of simulation Then the simulation can be carried out in a short time, on the
and testing is necessary for control design and verification seconds or minutes level.
[4]. To eliminate the testing time, an accurate and simple
engine model is expected to analyze engine dynamics in The rapid development of computer technology provides
order to design stable and robust control. The fact is that increasingly powerful capability for control engineering. So
control engineers have to trade off the model accuracy and far, a personal computer (PC) can afford for the integrated
the model computation time. simulation interacting between the 1D detail model and the
Simulink control model at the same time. Even though the
In the area of engine design and control design, different computing speed is still very slow for control design
engine models at different detail levels are available for purposes, but it is fast enough for control verification for both
different applications. The combustion chamber design needs steady and transient performance simulations. Compared with
to investigate the atomization property of fuel injection, the experimental data, the validated GT-Power model is capable
fluid turbulence, and the combustion characteristics, etc. The of predicting the engine performance with an error generally
KIVA [5] model is designed to describe three dimensional less than 3% [8]. The advantage of the integrated simulation
fluid dynamics and chemical reactions in the cylinder with is that it simultaneously simulates the engine performance
simulation time in the order of 10-100 hours. To study the and control function. It can be utilized for control algorithm
performance of an engine system including cylinder, verification and fault detection and diagnosis algorithm
turbocharger, intercooler, wastegate or variable nozzle verification. In this paper, the integrated simulation is used to
turbine (VNT), EGR, intake and exhaust manifolds, the 1D represent the test engine and is compared with the MVEM
detail simulation model can be developed using commercial results.
software tools such as Gamma Technology's GT-Power or
Ricardo's WAVE. The simulation time scale is in hours, Control design always takes several steps: modeling,
approximately 100 engine cycles. The GT-Power simulation simulation, HIL verification, and calibrations with a real
tool also provides control design modules, but they are plant. A general engine control design process is illustrated in
limited in flexibility and functionality. Figure 1. First, the 1D detail model can be started from the
prototype engine, or from the original model with adjustment
As the modern internal combustion engine becomes more and of the sub-systems. Second, performance simulation, like the
more complex to meet increasing demands for lower constant speed load acceptance (CSLA) test, the federal test
emissions and higher fuel economy, the development of an procedure (FTP) test, can be conducted to verify the design,
engine controller for such systems can be time-consuming emission requirement and subsystem configuration. During
and labor-intensive. Usually, this process includes many this phase, the controller can be included for operation point
iterations and extensive calibrations. As a result, control searching. The controller can be designed in GT-Power or
engineers prefer an accurate and simplified control oriented Simulink. During the second stage, the mean value model can
model to design and verify their algorithm, before applying be obtained based on the integrated simulation environment.
the results into on-line testing and calibration. The simplified More simulations need to be conducted to cover the wide
model should also capture the dynamics of interested range of engine operation conditions. Third, based on the
variables accurately enough. Meanwhile, most control mass conservation and energy conservation, each sub-system
engineers prefer Matlab and Simulink for their control module is modeled based on the simulation data regression or
design. Several mean value engine models were proposed to identification such as manifold volume, volumetric
meet such requirements. In [6], hybrid radial basis functions efficiency, etc. Then the sub- systems are integrated together
was used to approximate the simulation results of the detailed to form the MVEM. Fourth, the entire MVEM is verified
model for cylinder quantities. In [4], a Mean value model was with the 1D detail model in GT-Power. The MVEM captures
developed from standard experimental measurements the main dynamics of engine variables while there is error
(BMEP, VSFC, etc.) at partial and full load conditions at compared to the variables from the physical engine. Fifth,
different engine speeds. It was used to investigate the advanced control, such as robust control, is designed to cover
turbocharger lag in order to predict vehicle performance the model error and un-modeled dynamics. After stabilized
during the transient conditions. The advantage and control is obtained with the MVEM, a further verification is
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

necessary to check the control with the 1D detail model

again. Sixth, before testing the control on the prototype
engine, the hardware-in-loop simulations need to be carried
out. The design control can be downloaded to dSPACE or
Xpc, control system by interacting with GT-Power through
signal interface. Finally, the controller is connected to the real
test engine to verify the real-time capability, stability and
performance.

Figure 2. 1D detail model in GT-Power

Figure 1. Control design process for IC engine systems

Simulink based control was integrated with GT-Power for

In this paper, an integrated engine and control simulation this discussion, to facilitate the performance simulation,
environment is introduced in Section II. In Section III, the engine design, and MVEM development purposes. One
process to establish sub-system models is discussed. example is shown in Figure 3, to demonstrate how Matlab
Specified simulation needs to be performed to get data for controls the detailed engine model in the Simulink
regression. Later the integrated MVEM from sub-systems is environment. One advantage of the integrated simulation is
verified against the 1D detail model in GT-Power to that the feedback control can be applied to search operating
demonstrate the effectiveness of the developed model. To conditions in performance simulation [9]. In Figure 3, torque
conclude this method, the features and advantages of the control is used to find the full load fueling condition at
modeling are summarized in Section VI. different engine speeds.

INTEGRATED ENGINE AND

CONTROL SIMULATION
ENVIRONMENT
As shown in Figure 2, the 1D detail model is developed in
GT-Power for a 9L V6 turbocharged direction-injection
diesel engine. It is used as an example to demonstrate the
process to develop the MVEM. The engine specifications are
attached in the Appendix. The detail GT-Power model has
been validated against experiment measurements from a test
cell. The accuracy and computation speed always conflicts. Figure 3. Integrated 1D detail model with Simulink for
Because the 1D model is simplified for MVEM development, torque control.
the errors will be introduced. In fact, no model is perfect. The
model error and un-modeled dynamic will be considered in
the control design along with robustness and adaptivity. MEAN VALUE ENGINE MODULE
The control oriented sub-system models of a turbocharged
diesel engine system are presented in this section. The system
layout is shown in Figure 4. These sub-system models
include compressor, intercooler, intake manifold, engine
combustion, exhaust manifold, turbine, wastegate, engine
crank shaft dynamic, and turbocharger dynamic. The detail
process of modeling and verification of each module are
introduced. The model parameters are identified using
regression techniques. Each model was verified with
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Figure 4. Engine Structure and main variables of MVEM

Figure 5. Compressor performance and efficiency map and interpolation

simulation data from the 1D detail GT-Power model. With

the implementation of models in Simulink, each MVEM sub-
system is verified with the entire GT-Power engine model for
steady and transient state. (1)

For the mean values model, it was assumed that the air obeys The inlet of the compressor is assumed at ambient conditions.
the ideal gas law, the pressure is uniform in the intake and
exhaust manifold, and there is not any heat losses to the
walls. In combustion, heat is released in the whole
combustion chamber at homogeneous conditions, and the
gases can be regarded as ideal gases. (2)

To avoid using a complex function and regression, look-up

COMPRESSOR tables are used for the flow rate and efficiency prediction.
The turbine and compressor are connected through a shaft to The compressor map is extracted from the table defined by
utilize the kinetic energy from the exhaust to increase air operation data from the manufacturer. For other operation
density in the intake manifold. The mass flow rate through points in between, the interpolation is applied as shown in
them depends on the pressure ratio between the outlet and Figure 5 for both flow rate and efficiency.
inlet, and the corrected speed. They are modeled based on
map data from the manufacturer. The compressor exit temperature can be calculated by

As described in [10], the compressor's corrected mass flow

rate, efficiency and the temperature change are modeled as a
function of the pressure ratio and the corrected turbine speed.
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

(3)

where ηcomp is the isentropic efficiency of the compressor,

which is obtained from the look- up table. The γ is the
specific heat ratio, which is 1.401 at the intake air path.

INTERCOOLER Figure 6. Estimated intake manifold volume and the

The air temperature Tcomp increases after the compression, so mean value
the intercooler is used to cool it down. The gas temperature at
the exit of the intercooler Tboost is given by

(4)

where Tcoolant is assumed to be the same with Tamb for

simplicity, ε is the effectiveness of the intercooler. The
pressure drop, filling and emptying effect of the intercooler
volume is considered together with the intake manifold.

INTAKE MANIFOLD
The intake manifold is modeled as an open thermodynamic
volume, filling and emptying with an ideal gas by the
compressor and engine air charging process. Neglecting heat
losses through walls and assuming an ideal gas with constant
specific heat, intake manifold pressure is modeled as
Figure 7. Comparison of intake manifold dynamic model
prediction and GT-Power data

(5)
ENGINE MODEL
where pim and Tim are the pressure and temperature of the
Air-charging model
intake manifold, and R is the ideal gas constant. The
thermodynamic state ṗim is used to describe the results from Air charging into a cylinder is a highly nonlinear process
depending on the volumetric efficiency ηvol, engine speed Ne
the filling process of the compressor flow and the emptying
process of air sucking/changing into the cylinders. The and intake manifold states of Tim, pim, which describes the
manifold volume is the key parameter in the model. Based on engine pumping process as
the simulation data obtained over an operation range from
900-2100 rpm, Figure 6 shows the volumes and the mean
value. The intake manifold pressure dynamic model
prediction is compared with the GT-power model in Figure 7. (6)

where Vd is the displacement volume. The volumetric

efficiency is mainly a function of engine speed and intake
manifold pressure. Intake manifold change Δpim is included
to consider the dynamic effects of the air-charging process.
Back pressure effect is taken into account in terms of pem -
pim. Then the air-charging process is modeled as
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

(7)

The comparison between model prediction and GT-Power

data is shown in Figure 8. Both steady state and transient
processes are well predicted by the model.

Figure 9. Engine torque model prediction and GT-Power

data

EXHAUST MANIFOLD
To model the conditions in the exhaust manifold, the
temperature of the mass flow from the cylinder into the
Figure 8. Air-charging model prediction, GT-power data exhaust manifold is necessary. It is a function of fuel flow, air
and prediction error flow into the cylinders, and engine speed.

Engine shaft dynamics

From Newtonian Mechanics, the engine crankshaft dynamics
can be derived as in (8),
(11)

Model prediction and GT-Power data are compared in Figure

(8) 10.

where Ie is the engine inertia. TQb is the break torque load

and TQl is the torque load. TQb = TQind − Tf is the difference
between the indicated torque and the friction torque.

As the air charging is proportional to intake manifold

pressure [11], the simplified brake torque is modeled as

(9)

Fuel rate mf and engine speed Ne are obtained from offline

simulation. Linear regression result is as
Figure 10. Exhaust Temperature model prediction (net)
and GT-Power data (dots) comparison
(10)
Exhaust pressure dynamics are modeled as in the intake
The model prediction and GP-Power data are compared in manifold.
Figure 9.
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

through it can be described as follows, with supersonic and

chocked flows.

(12)
where ṁem is exhaust mass flow rate, which consists of fuel
and air flow rate into the cylinder; ṁturb and ṁwg are turbine
flow rate and wastegate flow rate respectively. The wastegate
opening is used to adjust the bypass flow ṁwg, in order to
control the exhaust manifold pressure. Engine air density is
indirectly adjusted through the turbocharger power by the (13)
exhaust manifold pressure. The estimated Equivalent exhaust
manifold volume and the mean value are shown in Figure 11. where CDf is the discharge coefficient of the orifice opening
One of the model predictions at 1100 rpm is illustrated in of max area Aorifice; p1 and p2 are inlet and out pressure.
Figure 12 along with the comparison of the GT-Power data.
For model identification and later verification purposes, a
GT-Power model was run at the throttle angle of 14.3
degrees, with the wastegate in closed condition. A transient
process of the turbine was obtained. The transient is from the
initial turbine speed of 90,000 rpm, slowing down to 80,000
rpm because of the latency of intake manifold pressure. Not
much air is available for combustion, as the stoichiometric is
maintained for the SCI scheme. Then gradually, as the
density of the intake manifold increases, the fuel injection is
increased, more exhaust power is available to speed up the
turbine. In this simulation, the turbine speed reaches
stabilization around 100,000 rpm. In the control analysis, this
Figure 11. Exhaust manifold volume and mean value non-minimum phase characteristic cause's difficulty in the
dynamic control and stability. The special phase latency
needs to be taken into account for the engine control design,
in which the system control is more complex than the control
in conventional diesel engines.

The unknown parameter of model (13), CDfAorifice, is treated

as one parameter in the model identification. The combined
effect of turbine dynamic change in speed and mass flow rate
is embedded in the integrated parameter. Notice that the
specific heat capacity ratio at the exhaust pipe is about 1.3,
and the critical pressure ratio is 0.546. As shown in Figure
13, the turbine exhaust flow runs through the equivalent
orifice under both subcritical chocked flow conditions.
Figure 12. Model prediction of exhaust manifold Separate models need to be considered.
pressure dynamic (1100rpm) and GT-Power data

TURBINE AND WASTEGATE

Turbine
Turbine performance is important for engine control. The
easy model, look-up table, is usually used for the turbine flow
model. Manufacturer provided flow map data and efficiency
data were used in the look-up table model first, and then
implemented in the Matlab 2D look-up table. The accuracy is
not enough in the steady state and transient state verification.
Then the turbine was modeled as an orifice. According to the
orifice model stated in throttle modeling; the mass flow
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Figure 13. Turbine fluid dynamic under subcritical and

chocked flow conditions

The detailed simulation data are shown for correlation

analysis among the flow rate, turbine speed, and discharge
coefficient and opening pressure ratio. From Figure 14 (a)
and (c), the equivalent turbine discharge coefficients have
different stages under supersonic and chocked flow
conditions. The correlation between the discharge coefficient
and turbine speed is obvious from Figure 14 (a) and (b). For Figure 14. Correlation between the discharge coefficient
the supersonic flow, the discharge coefficient has little of turbine and the turbine speed under subcritical (pr>
oscillation, but the trend follows the turbine speed 0.546) and chocked flow conditions (pr < 0.546)
development as it decreases and then increases. For chocked
flow, the mass flow rate will only depend on the upstream
pressure and temperature. But in this case, the down steam
pressure and temperature do not change dramatically and the
ambient pressure and temperature are adopted for
simplification.

According to the correlation between the discharge

coefficient and engine speed, two separate models are
created. The model of the product of discharge coefficient
and the equivalent opening of the turbine is obtained from the
GT-Power simulation data as a function of the turbine speed
as

(14)

Figure 15. Turbine discharge coefficient time equivalent

opening fitting model and comparisons with GT-Power
(15) data
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Figure 18. Turbine flow rate model in Simulink

Wastegate
Figure 16. Turbine flow comparison between GT-Power To simplify the wastegate flow-rate modeling, it was
data and model prediction modeled as an orifice as described in (15). The discharge
coefficient CDf_wg is estimated based on the linear model of
the wastegate opening Awg, and engine speed Ne,
Based on models (14) and (15), the orifice model based
turbine flow estimation was compared with the GT-Power
data in Figure 12(d). The comparison is shown in Figure 16, (15)
which has good consistency for both the subsonic and
chocked flow conditions during the entire GT-Power process. A comparison of discharge coefficient and flow rates from
the models and the GT-Power simulation are shown in Figure
In another set of simulation data, the turbine dynamic is 19 and Figure 20.
shown in Figure 17. The first three cases of 900 rpm, 1100
rpm and 1500 rpm, show good agreement of CDfAturb with
the GT-Power data. For the 1800 rpm and 2100 rpm cases,
the turbine runs out of map without control. It also implies
that the control design is important for the engine to run
within safe system operating conditions.

Figure 19. Discharge coefficient of wastegate as orifice

Figure 17. Comparison of equivalent turbine opening

and GT-Power data
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Because the exhaust temperature changes significantly over

the engine operating condition, specific heat needs to be
modeled.

(20)

Figure 21 shows the model prediction as a function of inlet

and outlet temperature of the turbine compared with the GT-
Power data.

Figure 20. Comparison of model prediction and GT-

Power data of mass flow rate through wastegate

The turbine exit temperature, Tt, is given by

(16)
where the turbine efficiency is determined by a look-up table
based on map data as a function of corrected turbine speed
and pressure ratio. Figure 21. Specific heat data and regression at turbine

MVEM IMPLEMENTATION IN
SIMULINK
(17) Each sub-system model is implemented in the Simulink
environment. They are integrated as a whole engine MVEM
The turbine outlet pressure is assumed as an ambient
to predict engine dynamics for engine system control and
condition for simplification.
power-train control design. The layout of the integrated
Simulink model is shown in Figure 22. The wastegate
Turbocharger dynamic opening control and fuel amount control will be designed for
The Turbocharger dynamics depend on the difference of engine operation. Furthermore, EGR and vehicle model may
power consumed by the compressor, Pcomp, and the power be included for after-treatment and power-train system design
generated by the turbine, Pturb. and simulation.

MVEM VERIFICATION
After the development of the mean value engine model, it is
(18) necessary to verify the effectiveness of the MVEM against
the detailed engine model data. The verifications are
where Itc, is shaft moment inertia 1.5×10−4 kg · m2; the
conducted in two steps: (1) Steady state tests of each unit and
powers are given by (2) a transient dynamic test of the whole system.

(19)
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Figure 22. MVEM of a diesel engine in Simulink

1. Steady State Verification: Unit Test for Given state. The engine exhuast temperture model follows the detail
Input model with little error in steady state. Fuel injection in (d)
were fixed with air mass flow rate, and has similar results
Each unit was driven with original data from the 1D detail
with (a). Engine speed is of the most concern in all speed
model, which were extracted from designated simulation
governing; it shows satisfied accuracy corresponding to the
runs. The given conditions are listed in Table 1.
GT-Power data. As indicated in Figure 27(a) and (b), the
exhaust manifold model results have a large error in transient
Table 1. External conditions for units verification and steady-state. It stabilized into a value of 2 bar. But the
exhaust temperature follows in good consistency with the
GT-Power in Figure 27(b). The turbinocharger speed has a
stable value of 95000 rpm corresponding to 100,000 rpm in
Figure 28. Shown in Figure 29(a), the turbine air follows the
combination of air change and fuel injection, the model data
are correct following the GT-Power in transient and steady
state. Finally, the turbine power output, which is important
For each of the units, the inputs were driven by the ideal data for speed estimation, has good accuracy in steady state as
from the detailed simulation results, just like the model part shown in Figure 29(b). The turbine efficiency differenc may
was embedded into the detail model. Figure 23 shows the be the cause of transient error.
compressor verificatoin results. It can be found by the mass
flow rate through the compressor in (a), compressor outlet Overall, each engine unit has similar behaviors to follow a
temperture in (b), and compressor power output in (c); all steady control variable of throttle; they stablize to the same or
have good consistency in the steady states and transient close to the value in the long run. Further verification of the
processes. The spines of the model curves at about 1s and 3s whole system with any given input will test the interaction
in the three figures may be caused by the discontinuity of between the units in steady state and transient.
interpolation of the look-up maps. The Intercooler pressure
dynamic follows the detail model curve very well, the steady
state is very close to the GT-Power data in Figure 24(a).
Intercooler temperture doesn't show a big fluctuation and the
model data is close enough to the ideal data in steady states in
Figure 24 (b). The air flow through the throttle is predicted by
the orifice model in Figure 25. The model prediction
flowrate's initial state difference is less than 1s, which can be
ignored; the overal prediction is fine. There exists little steady
state error and a drop during transient. For engine cylinder
model verification, the model simulation results are shown in
Figure 26 (a) to (d). In figure (a), the air charging has good
consistency to the GT-Power model, especially in steady
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Figure 25. Unit verification: throttle flow rate

Figure 23. Unit verification: Compressor (a) mass flow

rate, (b) temperature output, (c) power output

Figure 26. Unit verification: Engine, (a) cylinder flow

rate, (b) exhaust temperature, (c) Fuel injection, (d)
engine speed

Figure 24. Unit verification: Intercooler, (a) boost

pressure, (b) boost temperature
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Figure 27. Unit verification: Exhaust manifold (a)

pressure, Exhaust temperature
Figure 29. Unit verification: Turbine; (a) Turbine flow
rate; (b) Turbine power

2. Transient State Verification

The transient dynamic of the MVEM is critical for the system
model verification. Because of the inevitable model error in
the unit model, the serial interaction between each part and
the feedback effects from the turbocharger loop makes it
difficult to test the model for a given input. The verification
condition was defined based on the engine speed, which is
the most important variable for power-train control. Engine
inputs and other internal variables, manifold pressure and
flow-rates for instance, were investigated accordingly with
the GT-Power model. The simulation results and comparison
with the GT-Power data are listed in Figure 30, Figure 31,
Figure 32, Figure 33, Figure 34, Figure 35, Figure 36, Figure
Figure 28. Unit verification: Turbine Speed 37, Figure 38, Figure 39 as follows.

Throttle

Figure 30. Throttle opening for transient verification

 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Response in Engine Speed and Turbine Speed Compressor

Figure 31. Engine speed dynamic, (a) in GT-Power, (b)

in Model

Figure 33. Compressor dynamic transients, (a) flow-rate,

(b) temperature, (c) compressor power

Intercooler

Figure 32. Turbine dynamics, (a) in GT-Power, (b) in

MVEM

Figure 34. Intercooler transients in (a) pressure, (b)

temperature
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Engine Cylinder
Throttle

Figure 35. Throttle flow rate transient

Intake Manifold

Figure 37. Engine cylinder transient in (a) charging

flow-rate, (b)exhaust temperature, (c) fuel injection, (d)
engine speed

Exhaust Manifold
Figure 36. Intake manifold transients in (a) pressure, (b)
temperature

Figure 38. Exhaust manifold transient in (a) pressure,

(b) temperautre
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

Turbine difference in steady state from the GT-Power because of the

difference of upstream and downstream pressure from the
manifold in Figure 36(a). The cylinder charging in Figure
37(a) and fuel injection in Figure 37(c) also follow similar
transients as they are defined by the throttle flows. As
mentioned formerly, the engine speed follows the GT-Power
data's transient in time as shown in Figure 37(d). The exhaust
manifold variables in Figure 38 are the key for the turbine
control; the turbine speed in Figure 39(d) shows the
consistency of the MVEM in transient with the 1D detail
model.

The consistency of the MEVE in both steady state and

transient state gives it the capability to represent the engine
operation for control design. For control design purposes, the
state space model can be further extracted by using the
Simulink function.

CONCLUSIONS
A control oriented mean value engine modeling process from
the 1D detail model is discussed. The goal of this modeling
discussion is to demonstrate the feasibility of a new approach
for an engine model. The MVEM is established from
software simulation data instead of engine test data, so the
Figure 39. Turbine transients in (a) flow-rate, (b) power cost and time to develop the model is reduced. There is good
output, (c) speed. agreement between the modules and the detail model. The
main features of this novel modeling method are summarized
as follows.
Because of the steady state error existing in parts, the throttle
opening was chosen to get a similar engine response for the 1). It is a cost effective modeling method to keep sufficient
transient verification purpose. The opening in the MVEM accuracy while reducing the complexity. For control design
steps from 14.9 degree to 15 degree at 15 second, then back application, simplified models are appreciated because large
to 14.9 degree at 35 second, corresponding to the opening in amounts of testing simulations are needed to run during the
the GT-Power of 14 degree to 14.3 degree at 15 second, then initial design. Instead of running the engine in a test cell,
to 14 degree at 35 second in Figure 30. As shown in Figure detailed simulations are conducted to extract the dynamic of
31 and Figure 32, the major concerns in the power-train each engine sub-system. Risks and accidents could be
control are the engine speed and turbine speed. The transient avoided during the initial engine design process. Some detail
process of the MVEM has a settling time of 10 second in modes could be inherited from the former engine. Control
Figure 31(b), compared to the similar settling time of 13 development can be conducted with other designs
second. At the same time, the turbine dynamic shows a close simultaneously.
match in the transient characteristics and value. The steady
2). Computation speed is dramatically increased for the new
state value is not worth comparing so far, because of the
model. The new model consists of lumped parameters or
steady state error in parts. But the steady state error could be
simple nonlinear polynomials. No iterating computation for a
eliminated by the compensation in the efficiency parameters,
differential equation is needed in the MVEM based
for instance, in the turbine efficiency or volumetric
simulation. Computation time could substantially be saved by
efficiency.
using the MVEM.
The internal variables of the compressor are shown in Figure 3). Accuracy is slightly compromised due to the neglect of
33, the flow rate, temperature and power output follows the heat losses, approximations of temperature calculation over a
date from the GT-Power in transient and steady state. The wide range of operation points are compared with the detail
intercooler variable of pressure has an error of 0.1 bar out of model. However, the new MVEM still keeps the capability to
2.2 bar in steady state, but has a similar trend in the transient predict engine behavior for dynamics analysis of the complex
in Figure 34(a). The intercooler outlet temperature has an engine system. As the parameters go through a wide range of
identical value with the GT-Power data in Figure 34(b). The simulation data, all the operating conditions will be covered
throttle flow rate from the MVEM in Figure 35 has a in the new model too.
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

4). Each model of the engine module is reusable for other in Automotive Diesel Engines,” SAE Technical Paper
specific engines. Only the model parameters need to be 2007-01-1301, 2007, doi:10.4271/2007-01-1301.
identified again. They could be extended to a more detail
5. Los Alamos National Laboratory, “KIVA-II: A Computer
parameter lumped model if a more accurate prediction is
Program for Chemically Reactive Flows with Sprays,” 1989
necessary.
5). The new modeling method has a certain commercial 6. He, Y., and Lin, C.-C., “Development and Validation of a
value, as it can be included into a commercialized software Mean Value Engine Model for Integrated Engine and Control
package to enhance the software function. All the modeling System Simulation,” SAE Technical Paper 2007-01-1304,
data are already available once the detail model was 2007, doi:10.4271/2007-01-1304.
developed. After the operation parameters are defined, the 7. Schulten, P. J. M. and Stapersoma, D., “Mean Value
software could run the sub-routine to extract the MVEM Modelling of the Gas Exchange of a 4-stroke Diesel Engine
automatically. for Use in Powertrain Applications,” SAE Technical Paper
6). The trade-off of model accuracy and computation speed 2003-01-0219, 2003, doi:10.4271/2003-01-0219.
was explored and evaluated for the developed mean value 8. He, Y., “Development And Validation of A 1D Model of
model, compared with the 1D detail model. A Turbocharged V6 Diesel Engine Under Steady State and
7). The feasibility of a new modeling approach was Transient Conditions,” SAE Technical Paper 2005-01-3857,
investigated in detail. The control oriented models, like the 2005, doi:10.4271/2005-01-3857
state space model, the transfer function model can be 9. Wu, H., Wang, X., Winsor, R., and Baumgard, K.,
extracted further. The Simulink MVEM model can be easily “Integrated Simulation of Engine Performance and AFR
included in the control design environment. Control Control of a Stoichiometric Compression Ignition (SCI)
verification can be conducted with integrated simulation back Engine,” SAE Technical Paper 2011-01-0698, 2011. doi:
to the 1D detail model before conducting the HIL and engine 10.4271/2011-01-0698.
testing.
10. Heywood, J. B., “Internal Combustion Engines
8). The dynamic term is introduced and included in the
Fundamentals,” McGraw-Hill Book Company, New York,
volumetric efficiency model. Better transient accuracy is
ISBN 0-07-100499-8: 255-260, 1988.
obtained from comparison with the original data.
11. Hendricks, E., Chevaller, A., Jensen, M., Sorenson, S. C.,
ONGOING WORK FOR THIS PAPER Trumpy, D., and Asik, J., “Modeling of the Intake Manifold
Even though there is good agreement between each MVEM Filling Dynamics,” SAE Technical Paper 960037, 1996, doi:
module and the 1D detail model, there needs to be an 10.4271/960037.
integrated whole MVEM system testing for the wide range of
engine operating conditions for both steady state and transient CONTACT INFORMATION
state. Simple heat transfer and loss estimation will be
Hai Wu and Xinlei Wang are with the Department of
considered for the exhaust manifold, due to the high
Agricultural and Biological Engineering at the University of
temperature change over the speed range.
Illinois at Urbana-Champaign, Urbana, IL 61801 USA.
(Corresponding author: Hai Wu, 217-333-9415, fax:
REFERENCES 217-244-0323, e-mail: [email protected]). Richard
1. Bengtsson, J., Strandh, P., Johansson, R., Tunestal, P., and Winsor and Kirby Baumgard are with John Deere Power
Johansson, B., “Hybrid Modelling of Homogeneous Charge Systems, Waterloo, IA
Compression Ignition (HCCI) Engine Dynamics - a Survey,”
International Journal of Control, (80):1814-1847, 2007.
ACKNOWLEDGMENTS
2. Canova, M., Chiara, F., Flory, M., Midlam-Mohler, S.,
Guezennec, Y., and Rizzoni, G., “Dynamics and control of The authors would like to gratefully acknowledge the
DI and HCCI combustion in a multi-cylinder diesel engine,” colleagues from the John Deere Power Systems for their
presented at Fifth IFAC Symposium on Advances in support, which has been crucial to the success of this work.
Automotive Control 2007, Seascape Resort, California, USA,
2007.
3. Chauvin, J., Corde, G., Petit, N., and Rouchon, P.,
“Airpath Strategy for Experimental Transient Control of a
Diesel HCCI Engine,” Oil & Gas Science and Technology -
Rev. IFP, (62): 483-491, 2007.
4. Pettiti, M., Pilo, L., and Millo, F., “Development of a New
Mean Value Model for the Analysis of Turbolag Phenomena
 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

DEFINITIONS/ABBREVIATIONS Tamb [°K] Atmospheric temperature

Awg [mm] Wastegate opening Tboost [°K] Boost temperature at intercooler

Aorifice [mm] Orifice opening Tcomp [°K] Compressor outlet temperature

CDf Forward discharge coefficient Tex [°K] Exhaust temperature

CDf_chocked Forward discharge coefficient Tcoolant [°K] Coolant temperature

under chocked condition
Tim [°K] Intake manifold temperature
CDf_sub Forward discharge coefficient
under subcritical condition Tin [°K] Inlet temperature

Ie [kg·m2] Engine inertia Tturbine [°K] Temperature at turbine

mair [kg/s] Air charge into cylinder TQb [N · m] Engine break torque

ṁcomp [kg/s] Flow rate through compressor TQf [N · m] Engine friction torque

ṁcyl [kg/s] Flow rate into the cylinder TQind [N · m] Engine indicated torque

ṁem [kg/s] Flow rate into the exhaust TQl [N · m] Engine load torque
manifold
Vd [m3] Displaced cylinder volume
ṁim [kg/s] Flow rate into the intake
manifold Vim [m3] Intake manifold volume

ṁturb [kg/s] Flow rate through turbine ηcomp Compressor efficiency

ṁwg [kg/s] Flow rate through wastegate ηvol Volumetric efficiency

Ne [rpm] Engine Speed ηt Turbine efficiency

Ntc [rpm] Turbine Speed γ Specific heat ratio

pamb [bar] Ambient pressure ε Effectiveness of the intercooler

pboost [bar] Boost pressure

pcomp [bar] Compressor pressure

pem [bar] Exhaust manifold pressure

pex [bar] Exhaust pressure after turbine

pim [bar] Intake manifold pressure

pin [bar] Inlet pressure

pout [bar] Outlet pressure

R [kJ/kg × °K] Gas constant of the air

 Downloaded from SAE International by University of Liverpool, Sunday, August 12, 2018

APPENDIX
Engine geometry

The Engineering Meetings Board has approved this paper for publication. It has Positions and opinions advanced in this paper are those of the author(s) and not
successfully completed SAE's peer review process under the supervision of the session necessarily those of SAE. The author is solely responsible for the content of the paper.
organizer. This process requires a minimum of three (3) reviews by industry experts. SAE Customer Service:
Tel: 877-606-7323 (inside USA and Canada)
All rights reserved. No part of this publication may be reproduced, stored in a
Tel: 724-776-4970 (outside USA)
retrieval system, or transmitted, in any form or by any means, electronic, mechanical, Fax: 724-776-0790
photocopying, recording, or otherwise, without the prior written permission of SAE. Email: [email protected]
ISSN 0148-7191 SAE Web Address: http://www.sae.org
Printed in USA