You are on page 1of 18


discussions, stats, and author profiles for this publication at:

Decompression Engine Brake Modeling and

Design for Diesel Engine Application

Conference Paper · August 2010

DOI: 10.4271/2010-01-1531


2 1,525

4 authors, including:

Celso Argachoy
IMT - Instituto Mauá de Tecnologia & Cummins Inc.


Some of the authors of this publication are also working on these related projects:

Cellular Instabilities in Syngas/CH4 Mixtures View project

Conceptual study of an advanced ethanol-fueled engine View project

All content following this page was uploaded by Celso Argachoy on 17 May 2014.

The user has requested enhancement of the downloaded file.


Decompression Engine Brake Modeling and Design for Diesel

Engine Application
Frederico Augusto Alem Barbieri
Ederson Caludio Andreatta
Celso Argachoy
Hildebrando Brandão
MWM International Motores

Copyright © 2010 SAE International

The role of the engine brake is to convert a power-producing engine into a power-absorbing retarding
mechanism. Modern heavy-duty vehicles are usually equipped with a compression braking mechanism that
augments their braking capability and reduces the wear of the conventional friction brakes. This work presents
an engine brake mechanism modeling and design based on decompression effect, obtained by exhaust valve
opening during the end of the intake cycle. Besides that, during the system operation the emissions are
drastically reduced, even eliminated, since there is no fuelling, contributing to pollution level reductions. In this
sense, this work describes a development of such engine brake system for a 4 and a 6 cylinder diesel engines.

The engine brake performance was predicted by the development of 1D engine models. The 1D engine models
are able to simulate the valve train, including the valves operation, brake flap actuation, hydraulic actuator
behavior, and also the major engine breathing characteristics: gas flow rate, turbocharger efficiency,
temperatures and pressures along the intake/exhaust system, etc. The gas distribution along the exhaust system
can be predicted and its effects on the brake system performance evaluated. With these first assumptions, the
first prototype is constructed and the simulation results are compared to the test bench acquired data.

Conventional brake systems are designed to operate in very short periods of time. The continuous use of this
kind of system leads to excessive wear on its components and also causes dangerous overheating of braking
plates affecting driving confidence and security, mainly in trucks. Another issue related to conventional system
is fuel consumption, as the combustion process is kept activated during the brake operation. For long uses, such
as extended downhill routes, the amount of fuel burned and the brake components deterioration are
considerable. In order to preserve the conventional brake system and avoid unnecessary fuel burning, the engine
can be converted from a power-source to a power-absorbing retarding mechanism. Under usual operating
conditions, there are several sources of power dissipation, such as the pumping work (intake and exhaust
strokes), rubbing friction and accessories driving. The sum of these factors is the total friction power generated,

Page 1 of 17
which aids on engine braking [1,2]. However, to substitute the use of conventional brake systems, much more
braking power is necessary, requiring the development of auxiliary braking systems.

The engine brake converts a power producing engine in a power absorbing retarding mechanism and the
valvetrain is a significant part of this mechanism. All the engine brakes have one common characteristic: the
retarding horsepower increases as engine speeds increases. The major issues that affect the brake performance
are compression ratio, engine displacement, valve opening timing and the possible limits on turbo back pressure
and injector temperature.

This work presents the development of an engine brake with an exhaust valve opening system since its early
design stages, with 1D flow calculation, going through FEA and multibody simulation and also presenting the
first test results of a prototype. Figure 1 presents the methodology adopted.

FEA models –
of flexible bodies

Kinematic Dynamic multibody Geometry and

calculations simulation parameters

1D Flow Simulation –
gas exchange

Figure 1 – Engine Brake model simulations


Among several kinds of brake systems, the exhaust gases restriction is widely used, being the concept behind
this alternative based on power absorbing during the exhaust stroke.

Some of MWM International Acteon engines are equipped with the called Exhaust Brake System. In this
system, the exhaust gases are trapped in the engine by the use of an end flow flap installed after the turbine exit,
at the beginning of the exhaust pipes. When the system is activated, the flap is closed, avoiding exhausting. The
exhaust manifold is then fulfilled with pressurized gases, creating a retarding performance when the exhaust
valve opens and the piston is moving upwards forcing the intake air out of the cylinder.

The exhaust valve springs must be well designed in order to avoid valve bouncing due to high back pressure on
the exhaust runners. Continuous valve bouncing during brake activation may result in valve seat damages and in
some cases, valve breakage. A compromise relation between valve spring preload, flap leakage and maximum
pressure on exhaust pipes shall be achieved.

In the Exhaust Brake System, the ignition and fuel injection are turned off, reducing the amount of power
generated on the cycle. However, during the compression stroke, the pressurized air exerts a force on the piston,
returning most part of the power to the crankshaft.
Page 2 of 17
Keeping the exhaust valve opened during the compression stroke is an alternative to improve the performance
of the Exhaust Brake System. The energy stored in compressed air during the compression stroke is dissipated
in the exhaust system, even with the high back pressure caused by the exhaust flap, due to the exhaust valve
opening. During the expansion stroke, there is also a significant increase in braking efficiency as the piston
downwards movement generates a decompression on the cylinder chamber. The ideal exhaust valve lift to
achieve this operation condition is shown in Figure 2.

Figure 2 – Ideal Exhaust Valve Opening during Compression Stroke [3]

In order to achieve this braking behavior, it was developed by STEYR NUTZFAHRZEUGE AG a braking
system based on the positional locking of the exhaust valve, advantaging the fact that, when the exhaust flap is
shut, the high back pressure on exhaust runners causes valve bouncing. This is a simple and non-expensive
solution that can achieve improvements up to 65% of braking power when compared to the use of only an
exhaust flap. [3]

Figure 3 shows details of the mechanism developed to lock the valve open after bouncing. When the exhaust
flap is closed, the impulse wave of adjacent cylinders and the back pressure in the exhaust pipe causes high
back pressure values on the exhaust runner of a specific cylinder. When the valve bounces, moving down, it is
followed by the hydraulic actuator inside the rocker arm, driven by its internal spring. The volume in the rocker
arm chamber left by the actuator movement is then fulfilled with oil, which comes through the oil grooves along
the spherical check valve at the upper side of the actuator. When the exhaust valve starts closing, the oil
pressure inside the actuator chamber gets higher than the pressure in the line. The check valve closes and the
space with the oil above the actuator is fixed and the piston locks its position.

The exhaust valve will remain opened, with its lift is limited by the stroke defined by the actuator lateral rail,
which is locked by a pin. The high pressure of the oil at this moment forces the rocker arm upwards but the
rocker arm movement is avoided by the bracket above it.

When normal valve lift occurs, the rocker arm moves down, opening the oil groove that connects the oil
chamber to the top of the rocker arm. The actuator returns to its original position as the oil leaves through oil
groove. This sequence of events occurs every 720° of the crankshaft, while the brake flap is closed.

Page 3 of 17
Figure 3 – Exhaust Valve Brake System Scheme

The exhaust valve lift obtained with the system mentioned above is shown in Figure 4. During the expansion, as
the exhaust valve remains opened, the residual gases in the cylinder slight reduces the brake power efficiency
due to decompression.

Figure 4 –Exhaust Valve Opening during with Exhaust Valve Brake System actuation [3]

This system allows the valve bouncing but, as it is locking the valve opened, the detrimental consequences of
this effect, like seat wear and valve breakage, are avoided. Also, although the application engine has two
exhaust valves and the hydraulic actuator is installed in only one of them, it is quite not possible to occur valve
bouncing in the uncovered side. The bouncing is caused by the force balance between cylinder gas pressure,
exhaust runner pressure (back pressure) and valve spring pre-load. As the cylinder and back pressure
magnitudes are very similar for both valves, the actuator spring has a strong influence on the system confidence.
The actuator spring preload is about 10% to 20% of the valve spring preload, reducing the seating force on this
position, and guarantying that this valve opens with lower gas resultant force. As the pressure on runners
decreases, the other valve position remains closed.

Figure 5 shows a comparison between each braking system for the Acteon 6.12 V engine. The Exhaust Valve
Brake values were estimated based on a benchmarking comparison, considering base engine design parameters
that can directly affect the braking efficiency of this system, such as: displacement, compression ratio, boost
pressure, pressure before turbine, etc.

Page 4 of 17
The exhaust valve brake concept has been adapted by MWM International Motores (MIM), to be used in the
Acteon engine family, based on design and test results of other engine applications. Then, became the need of
knowing how suitable would be this engine brake system and how it would match with MIM applications.

The best way to get this kind of feedback to improve the engine brake design is analyzing its components
through FEA and 1D flow calculation, which give some outputs like reliability and performance response.

Figure 5 – Brake Performance Prediction for a 6 cylinder engine.

Studying the alternatives to predict not only the system operation and its components behavior but also the
influence of this operation on the engine performance, GT-Power v6.2 (1D flow simulation software) was
chosen instead of conventional CFD analysis.

The advantages of building a model using GT-Power are related to obtaining results that should aid during the
design step, avoiding waste time and eliminating costs with engine tests and prototypes construction. To
accomplish theses goals, the model must have high robustness and accuracy.

The first approach regarding the model was to correctly simulate the engine brake operation itself, like the valve
bouncing when the back pressure wave comes to the runner and the performance of the hydraulic actuator to
hold the valve opened. This can be done with a single cylinder model (Fig. 5), where the inputs are only the
boundary states just at the engine interfaces which provide results faster than a complete engine model. When a
good correlation for this first model is obtained it is ready to be coupled with the other cylinders and the whole
engine, where the increasing of brake power under different speeds can be analyzed.

The valve train components (mechanical parts of the system as valve and actuator) were modeled using linked
masses, and a ground template with an imposed movement was used for the rocker arm, as showed in Figure 6.
What must be pointed out is that the model elements at this stage have no flexibility avoiding deformation
during its operation. So the contact connections between the parts must consider the material properties and also

Page 5 of 17
the resultant stiffness of the designed assembly. The flexibility of some parts are then inserted in the multibody

Figure 6a – Single cylinder model

Figure 6b - Engine brake assembly.

Some design features, like spring pre-load, can be inserted into the model, providing mounting characteristics
that can be easily changed to evaluate their effect on the system. The contact model allows inserting different
multiplicity for each connection side, useful when modeling a hydraulic reduction mechanism or even multiple
elements linkage, for instance. The distance in which the stiffness and damping forces will be applied can also
be inserted, simulating an oil film between the parts or even a less rigid material. A non linear stiffness
coefficient, as shown in Fig. 7, is used to simulate a typical contact between the parts.

Page 6 of 17
Figure 7 – Contact stiffness transition.

Regarding the hydraulic mechanism, a dynamic valve connection was used to allow oil feed get into the
actuator and avoid leakage after bouncing. Input parameters like spring installed length and spring stiffness
shall be inserted in order to get the same dynamic performance as the real system. The leakage orifice,
representing the oil groove, was modeled to directly follow the clearance in the top of rocker arm. Its effective
diameter increases between zero and the nominal value (Eq. 1), in order to get the correct effect of fluid leaking
given by the rocker arm movement.

Deffective 2 Dorifice 2
Dorifice L  
4 4 (1)

The complete hydraulic system model is shown in Figure 8.

Figure 8 – Hydraulic system.

The forces evaluated in the model are used in the load balance on the exhaust valve (Figure 9). It is important to
note the influence of oil pressure and actuator spring forces that increase the occurrence of bouncing on the
valve with Engine Brake system. They also minimize the bouncing tendency on the other valve position due to
higher pressure acting downstream the valve (from the cylinder side).

Page 7 of 17
Figure 9 – Forces acting on the valve.

The resultant pressure force on both sides the valve must consider the exact diameter when the valve is fully
closed or fully opened (Figure 10). So, a smooth variation, from the minimum to the maximum valve diameter,
was done according to the valve lift.

Figure 10 – Valve reference diameters.

The valve displacement is then taken from the main model, which contains the default engine components, and
is inserted in the calculation process of the Engine Brake model. At every timestep, a feedback signal is
generated, sending back the valve mass displacement to evaluate all the gas dynamics.

In order to reach higher back pressures in the exhaust manifold, the flow flap located after turbine is closed,
avoiding the amount of burned gas flow. The opening angle of this exhaust flap must be adjusted at each engine
configuration, generating higher back pressures that match with the desired bouncing value, overheating limits
in the injector nozzle and turbine pressure limit.

The first outputs of the simulation were used to check the response of the Engine Brake model to mechanical
input parameters. The Acteon engine exhaust manifold is composed by two parts, each one related to a group of
cylinders (1 to 3 and 4 to 6). This configuration affects the pressure distribution along all cylinders ports, as a
group does not suffer the total influence of the other. Pressure waves were taken in the back side of the exhaust
valve (exhaust runners), so the geometry effect of the exhaust manifold can be seen in two wave groups, one of
high and one of low peaks. Figure 11 shows the back pressure values for cylinder1.

Page 8 of 17
Figure 11 – Back pressure on exhaust valve of cylinder 1.

The exhaust valve lift obtained for 2300 rpm is shown in Figure 12. During the compression cycle there is a
small drop on the lift value due to the high compression pressure inside the cylinder. Due to the surface area
ratio, the oil pressure inside the actuator is 10 or more times higher than the cylinder pressure (Figure 13). This
pressure then induces a reduction of the volume inside actuator chamber due to oil compressibility properties.
This effect can be avoided using higher oil feed pressures, as it is possible to faster fulfill the actuator chamber,
increasing its internal pressure. However, in some cases it cannot be done due to oil pump restrictions. Another
way is to increase the ball valve seat angle what also increases the discharge coefficient.

Figure 12 – Exhaust valve lift for a valve at cylinder 1.

Page 9 of 17
Figure 13 – Pressure inside actuator chamber.

Forces distribution on Engine Brake components were taken at the connections. Figure 14 shows the
contribution of stiffness and damping effects on the total contact force between valve and actuator that can be
further used as input data for structural analysis.

Figure 14 – Contact force between valve and actuator.

Page 10 of 17
To validate the model and check its accuracy and confidence some engine tests were done in order to get
enough data for comparison between numerical and experimental results. Some model parameters were chosen
as design variables, allowing the model calibration.

Although fully closed, the brake flap presented some flow leakage during engine tests, inherent from its design
construction. Also, due to high back pressure, the turbocharger also leakages, reducing the pressure level inside
the exhaust manifold. In the numerical model, the sum of engine local leakages was substituted by a unique
leakage, concentrated on the exhaust brake flap. Then, the discharge coefficient of this component was adjusted
to achieve the same pressure levels found on tests.

Another parameter considered for calibration was the turbocharger shaft speed. Due to the high amplitude of
pressure variation and to the turbocharger shaft inertia, the acceleration and deceleration effects on this
component were avoiding smooth results transition at each numerical cycle, causing a non-convergence of the
model. To facilitate the numerical solution, the turbocharger shaft initial speed was imposed during the initial
cycles, leading to a stabilized operating condition. After stabilization, the shaft speed is set free.

The imposed initial turbocharger shaft speed was defined by the bench tests measurements. During tests, this
speed value was in the range from 31000 to 32000 rpm for all tested conditions. In the numerical model, after
leaving the shaft speed free to be adjusted by the own engine operating conditions, it was observed that there
was any significant variation of this parameter. This fact shows that, eliminating numerical instabilities, the
model is representing with confidence the actual engine operation.

Figure 15 shows the comparison between measured and simulated results for the turbine inlet gas pressure (P3).
Figure 16 shows the same comparison but considering the achieved brake power. The difference between the
values is shown above the bars.

The simulated pressure and temperature values in other positions, like intake manifold, exhaust runners, etc., are
also matching with measured results, with deviations close to 4%.

Figure 15 – Model Validation - Turbine Inlet Pressure (P3) Comparison

Page 11 of 17
Figure 16 – Model Validation – Brake Power Comparison

Together with the 1D Engine Brake model, it was developed a multibody model [7] of the valvetrain in order to
perform some dynamic analysis with respect to gas forces obtained from the 1D Simulation.

With the dynamic analysis it was possible evaluate several important quantities of this valvetrain configuration.
The valves displacements, velocities and accelerations were evaluated for several engine speeds. It was also
possible to verify the contact forces on the between the various bodies, hertizian pressures and valve seating
velocities. These quantities were evaluated with the Engine Brake mode on and off.

Figure 16 shows the exhaust valvetrain. Its is possible to see the cam (1), the tappet (2), the push rod (3), the
exhaust rocker arms (4) and also the Engine Brake actuator (5) and its bracket (6) and also the valves and valves
spring. In the multibody model only the push rods and the rocker arms were not considered to be rigid. It is also
possible to see that the actuator only acts in one of the two exhaust valves as described earlier.







Figure 17 – Multibody Model

Page 12 of 17
The gas forces are considered to act in the valves as described in Figures 9 and 10. With the sweep analysis it
was possible to analyze the exhaust valve opening in various engine speeds (Figures 18 and 19) and it is
possible to correlate the increase in Engine Brake power (Figure 4) with respect to engine velocities.

Figure 18 – sweep analysis – exhaust valve lift in Engine brake mode.

Figure 18 shows the exhaust valve lift caused by the exhaust gas forces (bouncing) corresponding to the lower
values of valve lift on the left side of the curve. The higher values on the right side of the graphic correspond to
the normal lift of the valves caused by the camshaft.

Figure 19 shows the exhaust valve bouncing in detail. It is possible to see the higher values of bouncing for the
higher engine speeds. This result in gas exchange optimization and a better Engine Brake performance. It is also
possible to notice the higher overshoot bouncing values due to greater values of pressure at the exhaust runners.

Figure 19 – Detail of exhaust valve bouncing – sweep analysis

This situation is also encountered in the measured results for the valve bouncing. Figure 19 shows a measured
time table for the exhaust valve lift and the pressure variation right above the back of the exhaust valve. The
drop on the valve bouncing showed here is due to the pressure drop in the engine brake actuator, as mentioned
at the 1D MODEL DEVELOPMENT section. This effect was then fixed with higher values of oil supply
pressure and also fixing an oil leakage on the actuator prototype piece. The green and red curves were obtained
for different values of oil pressure supply. The Figure below shows higher lift overshoot values because the
considered engine speed was higher than the ones showed on Figures 18 and 19.

Page 13 of 17
10 10.0

9 9.0

8 8.0
7 su 7.0
6 (b 6.0

LIFT (mm)

5 5.0

4 4.0

3 3.0
2 (sec) 2.0

1 1.0

0 0.0





Figure 20 – Measured exhaust valve bouncing

It were also developed optimization studies with the objective to obtain valve spring values capable to increase
the braking power performance.

From literature and previous tests, it is known that the valve spring preload shall be the lowest as possible, so
the valve can bounce in reduced engine speeds, without requiring high back pressure values. However, it is also
known that, the higher the back pressure, higher the brake power capacity. The use of low spring preloads
promotes valve bounce at low speeds, with low back pressure values, but also reduces the brake power. On the
other side, high spring preloads requires higher back pressure for bouncing, increasing the brake power, but
only at high engine speeds.

Also, considering the same back pressure value, the lower the valve spring preload, higher the bouncing lift.
What contributes to the brake power increasing, as it easies the gas flow from the cylinder to the exhaust
manifold. [1,3,5]

A DOE study was then defined to evaluate what should be the optimized valve spring value that can define the
best compromise between valve bouncing and back pressure in order to increase breaking power. To continue
this evaluation, the focus was concentrated on the 2100 rpm condition, which presented the smallest deviation
from measured to simulation results. It was chosen the valve spring preload and the gas pressure before turbine
(P3) as the two independent variables to determine their influence on the system performance.

Although P3 was the objective of the comparison, it was necessary to consider the pressure value between the
turbine exit and the brake flap (P4) as the independent variable. P3 value must be an output of the model
because it is directly related with the pressure value along the whole engine air system (intake and exhaust).
Also, P3 shall not be imposed as a fixed variable for the DOE study as it is constant along the engine cycle. By
this way, P4 was considered as the variable. P4 values range was defined in order to achieve the same P3 values
found on tests, resulting in the same engine breathing.

To increase the exhaust pressure values (including both P3 and P4), it becomes necessary to reduce the leakage
value in the model. However, as the engine leakage calibrated based on the tests, a modeling contrivance was
used. The brake flap leakage was set constant, and non-feasible values of P4 were set in order to achieve the

Page 14 of 17
desired P3 values. This modeling approach results in less impact on the calibration variables and guarantees an
acceptable P3 behavior. Table 1 shows the value range for each independent variable.

Table 1 – DOE Independent Variable range

P4 (bar) Valve Spring Preload (N)

Baseline Value 3.9 350

DOE Range 1 - 300 200 - 400

Figure 21 shows the DOE map for P3 as a function of P4 and valve spring load. It can be identified two limit
curves. The first one represents the current P3 level found on engine tests (around 3.8 bar – curve 1). The
second curve is defined by the maximum pressure level supported by the turbocharger (5.0 bar – curve 2).

Figure 21– DOE Results – P3

These two curves are then projected on the Brake Power DOE map, defining the operational region that can be
achieved with the defined valve spring loads and P3 values (Figure 22). It is possible to define the optimized
spring preload that maximizes the brake power capacity (close to 280N – curve 3) at the current P3 limits. Also,
it can be observed that the maximum brake power is more influenced by the back pressure (indirectly defined
by P4 on DOE) than the valve spring preload.

Page 15 of 17
Figure 22– DOE Results – Brake Power

Considering the maximum pressure allowed by the turbocharger (5.0 bar), the DOE map shows that the
maximum braking power that can be achieved at 2100 rpm is around 174 cv. This value matches with the
benchmarking prediction, as it was previously presented in figure number 4.

The work presented a development of an Engine Brake based on the exhaust valve bouncing for the increase of
the retarding power of a commonly used Exhaust Brake. The design methodology included simulations with 1D
flow, FEA and multibody simulations.

The obtained results can confirm the good design assumptions done at the early stages of the Engine Brake
system development. Compared to the Exhaust Brake primary used, the Exhaust Valve Opening Brake System
improved the engine brake power in 52%.

From the numerical model and engine tests it can be concluded that the developed numerical model is robust
enough to predict with confidence not only the brake system performance but also its components behavior. The
first results comparison brings a good correlation of numerical and experimental results. The integration of the
various simulation models (1D, FEA and multibody) were able to predict the system behavior and diminish the
number of prototype parts and tests.

The DOE study was able to identify the optimized spring preload that improves the brake power capacity since
the back pressure magnitude has a strong influence on the engine brake performance. To improve the model
validation, it is planned to get some other engine tests results, such as valve lift diagram, oil pressure at
hydraulic actuator, etc. Bench tests are also recommended to evaluate the trends showed in the numerical

Page 16 of 17
[1] WANG, Y., Introduction to Engine Valvetrains, SAE International, Warrendale, 2007

[2] HEYWOOD, J .B., Internal Combustion Engines Fundamentals, McGraw Hill Book Co., 1988

[3] Schmitz, T. N., H. Bergmann and H. Daeuble: The New Mercedes-Benz Engine Brake with Decompression
Valve. SAE PAPER 920086

[4] EP0736672B1, Verfahren zum Motorbremsung mit einem 4-Takt-Verbrennungsmotor, 1998

[5] GT-POWER V.6.2 User’s Manual, Gamma Technologies, 2006

[6] HAAS, E., SCHLÖGL, H., et al. Ein neues Motorbremssystem für Nutzfahrzeuge, 1999.

[7] ADAMS Engine User's Manual, MSC/FEV, 2007.


We would like to acknowledge the support given by FEV on the development of the multibody valvetrain

We also especially acknowledge the great support and the modeling development done by Ivan Miguel
Trindade and Vinicius J. M. Peixoto, who were also responsible for this work development.

Page 17 of 17

View publication stats